Secure quantum-resilient smart city communication networks using QSC-Net with MF-MBO-based energy-aware task scheduling.

Secure quantum-resilient smart city communication networks using QSC-Net with MF-MBO-based energy-aware task scheduling - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Sci Rep . 2026 Mar 7;16:12534. doi: 10.1038/s41598-026-41015-2 Search in PMC Search in PubMed View in NLM Catalog Add to search Secure quantum-resilient smart city communication networks using QSC-Net with MF-MBO-based energy-aware task scheduling Nalavala Ramanjaneya Reddy Nalavala Ramanjaneya Reddy 1 Department of CSE, RGM College of Engineering and Technology (Autonomous), Nandyal, A.P 518501 India Find articles by Nalavala Ramanjaneya Reddy 1, ✉ , G Arul Dalton G Arul Dalton 2 Department of Computer Science and Engineering, Saveetha Engineering College (Autonomous), Chennai, India Find articles by G Arul Dalton 2 , K Swathi K Swathi 3 Department of CSE(CS), CVR College of Engineering, Hyderabad, Telangana India Find articles by K Swathi 3 , Dasaka V S S Subrahmanyam Dasaka V S S Subrahmanyam 4 Department of CSE, Avanthi Institute of Engineering and Technology (A), Affiliated to JNTUH,, Hyderabad, India Find articles by Dasaka V S S Subrahmanyam 4 , Lakshmi Kranthi G Lakshmi Kranthi G 5 Department of CSE, Aditya University, Surampalem, Andhra Pradesh India Find articles by Lakshmi Kranthi G 5 , Gujjeti Nagaraju Gujjeti Nagaraju 6 Department of CSE-AIML&IOT, VNR Vignana Jyothi Institute of Engineering and Technology, Hyderabad, India Find articles by Gujjeti Nagaraju 6 Author information Article notes Copyright and License information 1 Department of CSE, RGM College of Engineering and Technology (Autonomous), Nandyal, A.P 518501 India 2 Department of Computer Science and Engineering, Saveetha Engineering College (Autonomous), Chennai, India 3 Department of CSE(CS), CVR College of Engineering, Hyderabad, Telangana India 4 Department of CSE, Avanthi Institute of Engineering and Technology (A), Affiliated to JNTUH,, Hyderabad, India 5 Department of CSE, Aditya University, Surampalem, Andhra Pradesh India 6 Department of CSE-AIML&IOT, VNR Vignana Jyothi Institute of Engineering and Technology, Hyderabad, India ✉ Corresponding author. Received 2025 Oct 15; Accepted 2026 Feb 17; Collection date 2026. © The Author(s) 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ . PMC Copyright notice PMCID: PMC13087219  PMID: 41794855 Abstract Adaptable optimisation that preserves efficiency under time-varying system dynamics necessitates modern task management in innovative city development and in distributed edge–cloud computing systems. Conservative optimisation techniques such as genetic algorithms, particle swarm optimisation, and classical monarch butterfly optimisation (MBO) suffer from premature convergence, poor multi-objective performance, and limited adaptability to changing environments. Further, virtualised infrastructures contextualise operational constraints that impair their ability to homogenously support heterogeneity in task types and quality-of-service demands. We present a hybrid scheduling framework called multi-strategy fuzzy-enhanced monarch butterfly optimisation (MF-MBO) that combines fuzzy dominance for strong multi-objective ranking, self-adaptive quantum-inspired tunnelling (classical acceptance strategy) to escape stagnation, and bounded greedy migration for stable local refinement and load balancing. To accelerate convergence while maintaining task fairness across distributed virtual machines, MF-MBO dynamically balances exploration and exploitation. In the experimental evaluation under different workload conditions, MF-MBO clearly outperforms baseline algorithms, providing improvements of 17.4% in task execution time, 22.8% in load-balancing efficiency, and 15.6% in energy consumption. The results are reported with respect to the standard MBO, while we also compare them with both GA and PSO under the same evaluation budget and workload conditions. The results show increased operational efficiency and scalability, along with greater robustness across varying environments. The idea behind the introduced MF-MBO framework enables practical adaptation for smart city infrastructure services, distributed edge computing, and IoT-based applications, through a reproducible, explainable optimisation pipeline. The last part of this study reports empirical results and sets a few benchmarks to support future extensions, such as broader-angle benchmarking and hardware-aware validation. Keywords: Quantum-inspired optimisation, Fuzzy dominance, Task scheduling, Smart cities, Energy efficiency Subject terms: Engineering, Mathematics and computing Introduction As operations evolved and resources became limited, a need for complex decision-making systems emerged — motivating research into artificial intelligence, optimisation, and distributed systems. Orchestrate tasks across multiple agents in Edge computing systems in smart cities, disaster response networks, and IoT implementations whilst observing constraints on energy use, speed of processing and computational cost. Traditional optimisation techniques are insufficient because they cannot adapt their approach to explore complex solution spaces under uncertainty. Techniques from nature-based optimisation have been integrated with reinforcement learning methods, demonstrating strong potential to improve system value. However, the area is still plagued by slow convergence speed, high computational demand, and difficulties in dealing with many conflicting objectives 1 , 2 . The existing research demonstrates that genetic algorithms, particle swarm optimisation, and monarch butterfly optimisation (MBO) serve as metaheuristic methods for solving task scheduling and resource allocation problems 3 , 4 . The majority of these methods experience difficulties because they either reach solutions too early or cannot handle expanding task systems in changing environments. The research community has yet to fully explore fuzzy logic systems that combine multi-strategy dominance mechanisms with quantum-inspired search strategies for hybrid system development 5 , 6 . The implementation of quantum computing concepts for local minimum avoidance and solution diversity improvement via tunnelling and amplitude reweighting remains limited in practical, orchestrated systems. A novel method uses fuzzy logic, combined with multiple strategy enhancements and quantum-like behaviour, to perform task scheduling for distributed virtual machines. The research develops the Multi-Strategy Based Fuzzy Enhanced Monarch Butterfly Optimisation (MF-MBO) algorithm to optimise load equilibrium while distributing tasks and minimising energy usage across multiple agent systems. By utilising fuzzy dominance rules in search balance optimisation alongside self-adaptive tunnelling and greedy heuristics, the method achieves superior performance. This research makes several original contributions to present-day scholarship. Fuzzy enhanced sorting based MBO, and dominance based population refinement: These methods also complement the classic MBO by being able to rank the execution participation paths via a more accurate fuzzified path ranking scores. To escape local optima (and maintain genetic diversity over generations), the algorithm leverages a quantum-computation-based, self-adaptive tunnelling algorithm. We then demonstrate that a greedy migration strategy provides fast convergence times with little loss of solution quality. Optimisation algorithms serve as units for orchestrating large-scale distributed computing. This paper presents two main contributions: a novel fuzzy-logic task orchestration framework enhanced with multi-strategy modifications and quantum-like tunnelling for adaptive optimisation, and a comprehensive description of the MF-MBO algorithm implementation and evaluation within a simulated, innovative task distribution environment characterised by long-running tasks and changing load and virtual machine configurations. An extensive performance evaluation compares the MF-MBO algorithm with three baseline algorithms —standard MBO, PSO, and GA — in execution time, energy efficiency, and makespan. The algorithm is shown to be scalable and robust with respect to performance indicators across multiple experimental scenarios. Research Findings − Research findings show that hybrid optimisation applies to real-world distributed computing systems. While there have been significant advancements in post-quantum cryptography, QKD overlays, and federated security analytics, current smart city IoT protection is architecturally fragmented — encryption/key management, identity, routing protection, and anomaly detection are typically implemented as disjointed layers, resulting in mismatched and temporal trust states, high coordination latency, and poor end-to-end assurance across heterogeneous city services. In addition, quantum security in access networks is fundamentally link-dependent: While QKD offers strong secrecy, it is also vulnerable to link losses and environmental noise, and while PQC delivers stable performance across quantum and classical facilities, it does not leverage quantum-grade secrecy when link conditions are not too unfavourable. As a result, no “quantum-only” or even “PQC-only” defences can provide security, latency, and availability at city scale, all at once. Another disparity arises in terms of real-time adaptation and scalability: traffic, node mobility, and cross-domain interactions (e.g., transportation, energy, healthcare, public safety) are fast and dynamic, so security control, rather than being policy-static, needs to be continuously adaptive. This problem underscores the need for QSC-Net. This novel hybrid quantum-classical security approach combines link-aware hybrid encryption, AI-based adaptation, and distributed learning into a complete system that ensures confidentiality, integrity, and continuity of operations across various domains of smart-city IoT. Through four combined contributions, this work fills the above gaps by proposing QSC-Net, a cost-effective, unified, secure communication architecture for joint smart cities (1), and a hybrid QKD–PQC encryption layer that selects, as a function of link quality and operational constraints, quantum or post-quantum protection, thus increasing availability without compromising confidentiality. (2) a lightweight Q-PUF–aided trust and identity service that allows for fast and effective device-level authentication mechanism that was able to manage the rapid growth in the number of devices that some high-density deployments within the IoT can support; (3) an RL-based secure routing engine that performs on-demand secure routing decisions based on local trust dynamics and network variability, capable of real-time responsiveness to dynamic topologies; and (4) a federated anomaly detection layer to enable efficient sharing of cross-domain threat intelligence without centralizing sensitive data, which provides scalability and interoperability. Through extensive simulations, we show that QSC-Net offers a high packet delivery success ratio and scalability relative to representative baselines while maintaining low latency for trust establishment and attack response. Thus, MF-MBO in this architecture is QSC-Net’s internal scheduling and optimisation module for secure, energy-efficient, and scalable smart-city communication, and it does not operate as a stand-alone optimisation framework. The following section of this paper contains the organisation of its content. The review of optimisation methods for task scheduling and resource allocation with fuzzy logic and quantum-inspired computing approaches appears in Sect.  2: Related Work. The research relies on MBO and fuzzy dominance mechanisms, quantum optimisation principles, and their theoretical foundations, as described in Sect.  3: Preliminaries. The section provides a detailed description of the MF-MBO framework architecture, including its components and pseudocode, to show how its modules work together to optimise task orchestration. The section provides details on the dataset, simulation environment, performance metrics, and comparison results with baseline algorithms. The evaluation of results in Sect.  6 depends on the existing literature, yet it shows that the proposed method surpasses previous methods while highlighting its main limitations in Sect.  6.1. Section  7: Conclusion and Future Scope presents the research results and outlines future work, including testing in real-world environments and integrating hardware systems with quantum devices. The research framework allows for a straightforward presentation of the theoretical background, technical implementation, empirical evaluation, and broader significance of the proposed research. Related work Although research scientists have studied well-known heuristic, quantum-based, and artificial intelligence techniques for optimising intelligent cities, these designs still face several challenges. Danish Javeed et al. A novel security solution for Internet of Things (IoT) in 6G assisted by federated learning and quantum computing 1 . By reviewing existing barriers, it assesses technological evolution and its implications whilst suggesting a research agenda and a conceptual framework. An overview of the security challenges for 5G and 6G systems, Fadi Muheidat et al. Noting that quantum networking offers advantages in energy efficiency and security 2 , suggested further exploration of quantum-centric communication systems and described the security improvements enabled by quantum key distribution—research by Ching-Hsien Hsu et al. GCA-SC, proposed in 3 , runs in hybrid mode to minimise delay and energy consumption in smart cities. Simulation results yield a 39% delivery ratio, with 99 kbps of throughput per code@node average per system. Diksha Chawla, along with Pawan Singh Mehra, developed quantum-resistant solutions, such as QKD, based on their research into security threats to Internet of Things devices. The paper outlines future standards under development for quantum-safe IoT connectivity, discusses the impacts of quantum computing, and identifies related challenges. RevMahonShD, Muhammad Azeem Akbar et al. Notably, the authors of 5 present 6G quantum computing applications through a literature review and professional interviews, resulting in 49 best practices and 15 applications. While the underlying technology and QKD science would benefit from the directions emphasised in our research, the researchers found that much of the technology was already well established. Abeer Iftikhar et al. In 6 , an SLR on privacy, security, and trust risks and remedies in edge/cloud networks was conducted. It’s a review of the primary technologies, comparing pros and cons and recommending future studies dedicated to AI. Mohd Abdul Ahad et al. Here, we review the technologies facilitating the realisation of smart cities and propose nanotechnology and quantum computing solutions towards sustainability, while outlining environmental, socio-economic, and technological barriers and directions for research associated with local sustainability strategies 7 . Adarsh Kumar et al. CQS, IoQDs, QD: 8 real-time societal applications. The work details the advantages of a quantum-safe architecture with data-in-transit protection and leads to new applications and lines of research. The paper outlines the benefits of a quantum-secure architecture, including protection of building data in motion, and highlights future development opportunities for new applications and research. Zefeng Chen et al. Low et al. 9 explored aspects of Metaverse integration with smart cities, addressing privacy, interoperability, and ethical implications, showcasing innovative applications and foundational technologies, and calling for further interdisciplinary research. Zakria Qadir et al. Kikiras et al. 10 reviewed UAV path-planning algorithms for disaster response. They proposed an IoT-based UAV smart city system that detects algorithmic shortcomings and key issues and suggests hybrid cooperative solutions. Prabhat Ranjan Singh et al. 11 investigated IoT integration and AI role in 6G network development, investigating theoretical solutions in the context of smart cities for potentials(logical construction like low latency, etc.) and problems (unrealistic approach due to unresolved technical issue, etc.); Muhammad Waseem Akhtar et al. A study 12 analyzed a 6G network architecture by exploring tracks like development with artificial intelligence and quantum communication technologies, application areas, performance metrics, and future research challenges. Mehdi Hosseinzadeh et al. Sukumar et al. 13 presented a classification of IoT clustering for smart cities, reviewing 51 research articles and extracting performance indices (e.g., energy consumption and latency), advantages and drawbacks, and future directions. In the work by JawadAlia and MohammadHaseebZafar 14 , a new TCP-based MQTT system for smart cities was proposed that outperforms existing brokers in terms of latency and delivery performance. Future research will address incorporating a hybrid quality-of-service model with UDP. Minghao Wang et al. 15 recognised four main technological fields in the vision of 6G. However, they managed to find notes and documented only security vulnerabilities, along with basic knowledge probes that need further investigation. Zakria Qadir et al. Without revealing existing work and outcomes, for example, in 16 , the authors comparatively analysed 6G v/s 5G development mainly over THz and quantum communication technologies, AI-IoT integration and future research requirements for data-intensive intelligent IoE applications. Vita Santa Barletta 17 quantum-coupled SIEM with ML to develop an innovative city security solution in a hybrid quantum-classical system. The research outcomes show that threat identification methods are fast-paced, and future studies will focus more on improving data protection and security systems. Yifan Zhou and colleagues 18 analysed quantum algorithms to enable a range of power system applications, including not only immediate decision-making but also cybersecurity and grid optimisation. First, the end user has few well-standardised quantum programming protocols to leverage, which creates friction due to the lack of cross-domain experience among quantum computer users and marketers. Scientists believe the fundamental goal of this emerging field is to develop scalable quantum algorithms. The Cognitively Managed Multi-Level Authentication (CMMLA) system was proposed by Maryam Shabbir et al. 19 to protect the Nuclear Command and Control Centre (NCCC) by integrating cybersecurity, quantum computing, and cognitive computing technologies. In the experiment, FLQKD used high-security mode to achieve strong authentication and high throughput. In the future, these quantum computing systems will be available in real time, enabling scientists to continue their research. In work by Francesco Chiti and coworkers 20 , satellite and aerial platform operations were explored for global Quantum Internet (QI) quantum communication networks. The study investigates secure communication through QKD-based applications, the incorporation of SDN, and feasibility evaluation 21 . Moving forward, researchers will work to refine the systems used to deploy more satellites while making them more interoperable and quantum-enabled. Describing COVID-19-Tested BIC-Based CT Applications to Augment Smart City Infrastructure Arunmozhi Manimuthu et al. Arxiv 21 They clarify issues related to CT deployment, find the technology useful in urban applications, and highlight it as an area of further research. Researchers Lilia Tightiza et al. 22 suggested deploying virtual twins for real-time sensor monitoring and incorporating the metaverse into innovative grid systems for superior operational performance. Through their investigation, the researchers demonstrate the benefits and challenges this field offers and identify avenues for future research. In studying edge computing in the context of quantum and IoT computing, we analyse the progress in computing over six decades, as reported by researchers Sukhpal Singh Gill et al. 23 . This research focuses on existing decentralised system frameworks, the problems with their implementation, and future directions for research on these systems. Ahmed G. Gad et al. Starting with a WoS specific review of Blockchain 24 , provides a literature review from 2013 until 2020, including research priorities and challenges. The implications of this study guide the research direction, as they require further attention from researchers regarding scalability, security, and standards development. Sri Nikhil Gupta Gourisetti and others. Finally, an innovative DLT cybersecurity stack for smartgrids was introduced and analysed based on its performance in power use cases and how these use cases could be mapped to the OSI/TCP/IP models 25 . This system evaluation should highlight the current scalability limitations and security gaps, outline the system’s standard weaknesses, and outline the features targeted for future growth. Wenzhong Shi et al. 26 More specifically, urban informatics was a new technology for assisting in planning and transportation network tasks (GeoAI), and other existing technologies were reviewed from an integration perspective (in planning). Future trends in ethics & inclusion were opened for further research. Combined with blockchain and post-quantum encryption, the healthcare digital twin system proposed by Haibo Yi 27 incorporates cloud computing functions while enhancing security. It works at a higher level but does not measure user experience, ethics, or adherence to guidelines. Agostino Marengo 28 has provided a comprehensive survey of the use of AIs in IoT systems, highlighting privacy protection solutions and advanced data processing methods. What the findings say about ethical AI systems’ performance, highlighting their shortcomings, and areas that need more investigation. Hamlet Reynoso Vanderhorst et al. These include Chakraborty et al. 29 , who developed a framework called Meta Smart Twin Cities and studied the convergence of digital twin, IoT, BIM, blockchain, and the metaverse in the context of construction. Results are encouraging, with some ethical issues that require further research. Ching-Hsien Hsu et al. GCA-SC 30 is a green communication approach that injects a hybrid algorithm-based technology into smart cities to minimise latency and power consumption costs. Simulation-level testing of the system demonstrated improvements in delivery performance, throughput, and energy efficiency. Vita Santa Barletta et al. A Hybrid Quantum-Classical Architecture to enhance Smart City Security through QML and SIEM 31 . The first QBoost (when tested with QRadar and DLQC) was found to speed up threat detection by 70% compared to the baseline. This study will follow on from the transmission of secure data using these two approaches, implementing Quantum Key Distribution in combination with sophisticated integrated QML techniques. Haydar Yalcin et al. Ali O. Endaya H. Shakry E. M., and Hossom A 32 . we filled the gap in the scientometric study of supercomputing and quantum computing focused on cybersecurity research. This research also highlights existing gaps, nascent directions, and major players; validation by experts, together with technology roadmapping, will be the next step. Using real-time data on startups, the team led by Francisco Trincado-Munoz 33 analysed digital transformation in Advanced Producer Services. The new research uses European cities as a basis for understanding tech differences among world cities, although further research will be needed to test the findings. Wasyihun Sema Admass et al. 34 In their review of “Current Cybersecurity Trends, Issues and Future Perspective”, AI/ML and quantum-secure encryption were mentioned. It monitors real-time funding and collaboration, identifies dangers, and proposes tech solutions. His team built a framework based on IOTA’s tangle technology to establish standards for smart cities using DLT, enabling them to interoperate securely. Indeed, the research suggests positive results, but it is based solely on IOTA technology, while the following study will assess various Distributed Ledger Technology platforms. Bokolo, A.J. et al. 35 issue related to integration challenges and (or) a Meta Smart Twin Cities environment can tacklewith a fully integrated Meta Smart Twin Cities frame workthat a multi-dimensional Meta Smart Twin Cities (uses a (or) a Meta Smart Twin Cities environmentthat combines blockchain, IoT, digital twins, and metaverse. The system demonstrates potential to drive generational improvements in cities through theory-driven assessments of operational efficiencies. The integration of a servicing system raises many issues, including ethical and legal concerns. Simone Fischer-Hübner et al. 36 . Stakeholder interviews for the study were conducted in network and supply chain organisations across European critical sectors: 37 While the results will need to be verified in future studies, they form a foundation for future cybersecurity studies in the EU. Zefeng Chen et al. In particular 38 , discussed the integration of Metaverse technologies and smart cities, focusing on cross-disciplinary collaboration, and provided research implications, suggestions for future work, and elaborations on key advantages, critical technologies, challenges, and ethical aspects. The Emerging Drone Technologies (QD, IoT, and CQS) as explained in (Adarsh Kumar et al. 39 . It highlights social quantum computing opportunities, secure quantum benefits, trust issues, as trustless solutions, interoperability challenges, and proposes further research on quantum-safe system architectures. Marc Schmitt 40 provides an AI-based solution that identifies moving implications and security threats across the overall network, mobile networks, and the Internet of Things, using local intrusion-detection and virus-detection methods. Our evaluation framework demonstrates that ML models yield better performance but struggle with integration and need improved development for AI security use cases. Prabhat Ranjan Singh et al. 41 This paper describes the evolution of network technology, with a particular focus on AI functions in 6G networks. The study examines how the Internet of Things works, discusses its technological challenges, and explains that future networks must deliver high-speed, reliable performance with the lowest possible latency. M. Manulis et al. On the other hand 42 , examined satellite security vulnerabilities by analysing historical attacks and discussing threats to space infrastructure arising from satellite constellations. The following are upcoming satellite communication topics, issues, security concerns, and satellite technology developments mentioned in the document “Satellite Communication PEAS.” Abderahman Rejeb et al. In this work 43 , used Latent Dirichlet Allocation (LDA) on 4455 journal articles that integrate blockchain with the Internet of Things. The research examines security and operational issues and identifies focal points that warrant further study. Abdul Ahad et al. proposed a five-dimensional, attack-secure, component-based, innovative healthcare system over 5G 44 . This paper provides a holistic analysis of 5G healthcare networks, covering existing challenges, future research directions, enabling technologies, security issues, and defence mechanisms. Vita Santa Barletta et al. In one such contribution 45 , proposed a combined security solution integrating a SIEM with Quantum Machine Learning and developed a Hybrid Quantum-Classical Architecture to protect Smart City systems. The new system operates at much higher speeds and more efficiently, and researchers will now focus on further increasing data transfer speeds and the system’s security in their future work. Ahmed M. Abbas et al. An enhanced NB-IoT uplink scheduler to maximise spectrum usage and minimise smart meter signalling load was proposed in 46 . Results: There was 17.47% more spectrum used. The forthcoming research will be on better traffic control systems. DURR-E-SHAHWAR et al. In 47 , quantum cryptography research was examined for demonstrating how this technique can strengthen network security in the coming years. This work outlines current issues, remedies, and future research avenues for the functionality of quantum key distribution (QKD) systems. Singleton: A Lightweight Security Protocol for IoT Devices using Cryptographic Ratchets by Amir Hassani Karbasi and Siyamak Shahpasand 48 . End-To-End Encryption: The new digital currency platform is implementing end-to-end encryption while remaining scalable and compatible with existing protocols, and believes it can build its quantum-safe solution to run alongside its platform. DSM is a post-quantum cryptography algorithm developed by Swati Kumari and her team to defend IoT systems against Mirai botnet threats. The overall operating frequency, physical size, or data protection capacity of the system is improved, leading to lower power consumption, reduced latency, and error-free operations, thereby enhancing system security. Sanjeev Kumar Gupta et al. 49 and Sanjaikanth E Vadakkethil Somanathan Pilla and Kiran Polimetla 50 have reviewed quantum cryptography exploits to bolster the network against emerging quantum threats. This helps secure communication and provides much better protection than standard encryption systems. Amoldeep Singh et al. From Introduction to the Quantum Internet (QI) and its fundamental principles, technological components, and security features 51 . It describes elementary theories of quantum communication and discusses how decoherence affects systems, as well as the limitations we must overcome to facilitate future interdisciplinary research aimed at implementing QI. Arunkumar Muniswamy and R. Rathi 52 have introduced machine learning (ML) for threat detection and also consider overall IoT security in smart cities. Identification of future development paths: Analysis of machine learning algorithms, attack types, datasets, and defensive systems. Jaime Señor et al. 53 evaluated NTRU post-quantum cryptography on IoT devices using the Contiki-NG OS. NTRU, the study finds, works well with modern microcontrollers but is not as easy to run on older ones. Hardware acceleration offers better performance as researchers continue to expand the capabilities of various devices to support many Internet of Things (IoT) devices. Yuan Cao et al. In this survey, 5–7 architectures, levels, and protocols of QKD networks are presented 54 . It describes current challenges, demonstrates their progression towards future security functionalities, and motivates multidisciplinary research to enable Qinternet in its entirety. Geetanjali Rathee et al. For smart cities 55 , created an e-voting system based on the Internet of Things and blockchain, which focuses on using trust computation to identify intrusions into the network. Compared to baseline methods, the system performs better in security testing but still needs further validation. VINAY RISHIWAL et al. A case study was performed to validate the blockchain-based V2X communication system designed by 56 to facilitate secure smart-city data exchange. Development of AI, along with future 5G technology (and standardisation), will enhance the integrity of the system and user trust. Ying Loong Lee et al. Chai et al. 57 have also explored AI performance optimisation by defining critical technologies that will influence next-generation networks (e.g., 6G). Researchers need to ensure the system’s performance under multiple conditions, as it is highly energy-efficient, scalable, and adaptive, designed to work effectively. Marieh Talebkhah et al. using big data applications for analytics, architectural design, and decision-making 58 , examined practical innovative city development. The report highlights existing issues, outlines potential solutions, and identifies areas for further research. H. Szymanski 59 developed a “Cybersecurity via Determinism” paradigm, which aimed at achieving the security of deterministic IoT via D-switches and SDN. During SD-WAN testing, it provided robust quantum-safe protection, achieved high-speed operation, and required a negligible memory footprint. It offers reasonable cost minimisation and DDoS protection, but the lengthy setup process makes it difficult to use. The following initiatives will seek to improve real deployment and capacity to scale the systems. Mehdi Hosseinzadeh et al. This classification system for smart city IoT clustering methods, developed by Zareapoor et al. (2020), included 51 research articles between 2017 and 2021 60 . Results show load balancing benefits, system scalability, and power efficiency, while ambitiously aiming to extend its area of operation before implementation in the real world in future investigations. Maxim Kalinin and Vasiliy Krundyshev 61 presented techniques of quantum machine learning (QML) for intrusion detection that outperform classical machine learning by 98% accuracy and achieve higher speed on extensive datasets. Developing optimisation methods for quantum algorithms will be a focus of future scientific exploration. Studies by Deepa Pavithran and others 62 analysed the primary building blocks and the challenges and limitations of Internet of Things systems that implement blockchain technology. This research shows that Hyperledger Fabric outperforms IoT blockchains in practice. The next series of research would target storage system issues and user privacy issues. Description: Anca Jurcut et al. 63 . Investigated IoT security by mapping different threats and proposing an approach to mitigate them. This study presents the potential applications of quantum computing in IoT security while highlighting the importance of integrating a security-by-design approach and of collaboration among participants in the IoT ecosystem. Trung Q. Duong et al. 64 explored how the integration of quantum computing and machine learning could enhance 6G data management and security mechanisms. Quantum-inspired machine learning applications and their potential for operation in 6G networks are illustrated, along with the system constraints and research opportunities. Evangelia Konstantopoulou et al. Multimode architectures for the ZUC-256 and SNOW-V stream cyphers that are efficient in both performance and area were proposed in 65 . The system has been tested with FPGA and ASIC implementations that outperform the sensitivities of stand-alone cyphers, achieving high throughput and operational efficiency. Yuhao Bai et al. 66 have presented a blockchain-based hybrid architecture, effective verification mechanisms, and an incentive system to sustain smart city infrastructure for public participation consortia. It has been demonstrated to be accurate and efficient in simulations. Stephen Diadamo et al. In 67 , we propose a method for distributing the variational quantum eigensolver (VQE) across remote quantum machines. The results suggest that larger runtime trade-offs on Ansatz states. Next steps in optimising and scaling ManalM. The fourth multi-component system, known as the IBFO-ODLAD method for IoT anomaly detection, includes BOA for hyperparameter tuning, MLSTM for prediction, IBFO for feature selection, and Z-score normalisation 68 . The accuracy is 98.89%. Feature reduction is being investigated for further work. Muhammadasgharkhan et al. To address this issue 69 , examined the privacy and security challenges of UAVs, noting that standard cryptography has become vulnerable to quantum computing. The article provides an overview of Post-Quantum Cryptography (PQC) for UAVs, including its benefits, issues, and future research prospects. The advancement of the Differential crossover quantum PSO method for closer IoT applications is conveyed in this paper by Sheetal N. Ghorpade et al. 70 . It also achieves 25% lower location error than existing algorithms, improving localisation accuracy and convergence speed. Esmot Ara Tuli and his team researched the integration of quantum computing with the metaverse, assessing its feasibility and the middleware and application domains 71 . The document presents current challenges, unaddressed research areas, and potential solutions for developing a quantum-enabled metaverse. Alireza Shamsoshoara et al. 72 examined IoT security issues and fixes, with a particular emphasis on PUF-based key generation for authentication. The article presents the current advantages and disadvantages of PUF error-correction methods, along with upcoming research directions to improve these techniques. Sukhpal Singh Gill and Rajkumar Buyya 73 reviewed the evolution, current applications, and prospects of quantum computing and discussed its impacts on future systems. Content explaining quantum tech disruption across industries and software tools for quantum programming. The direction of the research will live on. Laura Saez Ortuno et al. Zhang et al. 74 investigated quantum computing applications in market research, focusing on its impacts on data security, instant simulation, market segmentation, and consumer behaviour analysis. The text covers implementation issues and potential follow-up research directions. ROZHIN ESKANDARPOUR et al. To see how quantum computing can improve grid optimisation and power system analysis 75 , carried out the research. The next step of the study has proved its promise in the real-world demonstration of the most practical complex power system problems with this system. Zebo Yang et al. 76 addressed the risks posed by quantum computing to blockchain security by contrasting post-quantum and quantum blockchains. It examines existing approaches, difficulties, and upcoming directions in the study of quantum-resistant blockchains. Adlin Sheeba et al. 77 present the SMO-SSODBN architecture in this study for effective web service classification in Internet of Things applications. With greater accuracy, lower latency, and enhanced security for innovative city services, BDN outperforms current techniques. Ramesh Chandra Panda and Md. Safikul Islam 78 presents an IoT-based wind-powered organic composting equipment for sustainable farming in the paper. Future research in a broader range of waste types increases productivity, reduces dependence on chemicals, and uses renewable energy. Hemant B. Mahajan et al. 79 offered blockchain-based approaches to Healthcare 4.0 security for Electronic Health Records (EHRs). It assesses different approaches, identifies areas requiring further research, and recommends ways to store and share data securely in the future. Cher Chye Lee et al. 80 evaluated threat modelling approaches for quantum computing hazards, focusing on PASTA for Cyber-Physical Systems. It highlights resource constraints, offers mitigation strategies, and identifies weaknesses to address quantum threats. A recent research study on green computing lays the groundwork for a discussion of the energy–performance trade-offs that hybrid, secure smart-city systems need to balance. Li et al. specify AFED-EF and sequentially prove that, by jointly optimising the efficiency and placement of multiple resources for energy-aware VM allocation of IoT workloads, power consumption can be reduced while maintaining reasonable service performance in cloud data centres, thereby emphasising the importance of co-optimising security-aware resource allocations 81 . Zhang et al. Motivate ECMS for ME and control adaptive network energy management to improve energy efficiency under dynamic MEC conditions, which closely parallels link- and context-aware operation in hybrid architectures 82 . Kumar et al. Build on this direction to offloading and scheduling integrated edge–cloud, taking into account the trade-off between energy savings and latency, and execution constraints, as it also reflects the real-life scenario in which security mechanisms impose overhead too and should be provisioned together with quality of service targets 83 . Finally, Xu et al. propose service-level and energy-aware scheduling in cloud data centres and demonstrate that SLA-driven objectives can be jointly optimised with power/cost objectives, thereby underpinning the thesis that secure communication designs should explicitly measure energy overheads and justify these trade-offs using scheduling-aware analysis 84 . In recent years, the exploration–exploitation balance has been analysed, and the results suggest that the Mountain Gazelle Optimiser could be further developed 85 . Learning-based system optimisation is gaining traction in recent IoT research. Khoshvaght et al. 86 present a multi-objective DRL framework that jointly optimises spatio-temporal latency in mobile IoTedge networks, demonstrating the benefits of adaptive decision-making under dynamic workloads. Complementarily, Sahin et al. introduce a Collaborative Intelligence paradigm to perform split-point computing on-the-fly for multi-task learning, effectively sharing the workload between the edge and the cloud 87 . The significance of adaptive intelligence and system-level coordination is thus inferred from these studies, leading to the development of integrated architectures, such as QSC-Net, that unify learning, optimisation, and secure communication mechanisms. In recent years, it has been proposed to explore learning-based and AI-driven approaches, such as reinforcement learning and deep neural networks 88 , for dynamic cloud task scheduling. Despite their high adaptability to changing workload conditions, these methods typically require a large amount of training data, leading to high computational cost, for which the need to continue optimising even lightweight metaheuristics remains relevant. The literature has also recently drawn attention to multi-objective task offloading and to the exploration of hybrid bio-inspired optimisation in cloud and edge computing environments. Li et al. Than et al. 89 developed a deadline-constrained task offloading framework using biogeography-based optimisation (BBO) in mobile edge computing, where the authors demonstrate better trade-offs among latency, energy, and execution reliability. Their approach is practical when there is a fixed deadline, but it focuses only on offloading decisions. It does not handle adaptive load balancing or dynamic virtual machine heterogeneity. More recently, Li et al. Results: 90 presented a hybrid bio-inspired algorithm that integrates edge-cloud computation, enabling energy-efficient multi-evolutionary search strategies jointly optimising for energy and execution performance. While it shows the advantages of hybridisation in distributed environments, this procedure assumes that the optimisation objectives are relatively stable and does not account for fuzzy dominance or adaptive exploration–exploitation control. On the contrary, the proposed MF-MBO framework improves hybrid bio-inspired optimisation by embedding fuzzy dominance ranking and self-adaptive search strategies to achieve reliable multi-objective scheduling under dynamically changing cloud and smart-city workloads. In Table 1 , we provide a comprehensive summary of prior work on quantum-aware security, Federated Learning, and blockchain-based authentication systems in smart city and IoT Ecosystems. Through its methodologies and use cases, the research shows critical improvements and reveals limitations. By design (i.e., data protection, decentralised trust, and computational performance), it identifies three key issues: a lack of quantum cryptography integration with federated systems, limited fast threat detection, and a system architecture that does not support fast-changing environments. These limitations highlight that the QFed-ChainAI framework, which motivates QFed-ChainAI as a seamless integration of quantum key sharing, federated model privacy, and blockchain auditing, is the need of the hour for secure and decentralised communication in future smart cities. Table 1. Summary of related work on quantum-aware and federated security approaches in smart city and IoT systems. Ref. Author(s) Focus area Key techniques / frameworks Contributions Research gap identified 1 Javeed et al. (2024) Quantum + FL + 6G for IoT Security Quantum Computing, Federated Learning, 6G Proposed an integrated framework for future-proof IoT networks Lacks real-time scheduling and energy-aware adaptability 3 Hsu et al. (2022, 2024) Green communication in smart cities Hybrid Algorithms (GCA-SC), delay optimisation Improved delivery ratio and energy efficiency in simulations No integration of explainable or quantum-inspired optimisation 4 Chawla & Mehra (2023) Quantum-safe IoT security QKD, Quantum Risk Assessment Reviewed IoT threats and post-quantum connectivity strategies No practical routing or task orchestration framework 17 Barletta et al. (2024) Hybrid quantum architectures for urban security QML, SIEM integration Boosted detection speed; hybrid quantum-classical framework Security-focused only; lacks energy-optimised scheduling 19 Shabbir et al. (2022) Quantum-based authentication for command centres FLQKD, CMMLA Demonstrated high throughput and authentication success Not generalizable for smart city or task orchestration domains 25 Gourisetti et al. (2021) DLT security for smart energy grids DLT Cybersecurity Stack OSI-mapped cybersecurity layers for energy applications Focused on static energy systems, not dynamic task orchestration 47 Durr-e-Shahwar et al. Quantum Cryptography and QKD Systems Systematic Review on QKD Comprehensive survey on QKD advantages and implementation directions Purely theoretical; lacks task-level scheduling or operational context 61 Kalinin & Krundyshev Quantum Machine Learning for Intrusion Detection QML, Quantum-enhanced ML High-accuracy IDS with quantum advantage No energy optimisation or task allocation strategy Open in a new tab Preliminaries The following section presents the theoretical foundations of the proposed system. The section begins by explaining the basics of quantum communication, then describes the security problems that arise in intelligent city networks and how federated learning protects data privacy during anomaly detection. Quantum-inspired sustainable AI model development requires these theoretical and mathematical foundations. Fundamentals of quantum communication and QKD Quantum communication relies on quantum-mechanical principles to build a secure information transmission system that cannot be broken. Quantum systems perform encryption using their special quantum properties of superposition and entanglement, which cannot be replicated or altered during measurement, unlike classical systems that rely on computational complexity for encryption. Quantum key distribution (QKD) serves as the fundamental component of quantum communication because it generates random secret keys between two parties while detecting any unauthorised access to their vulnerable communication channel. The BB84 protocol is the first QKD protocol and remains the most popular method since its introduction by Bennett and Brassard. The scheme requires quantum bits (qubits) to travel between two points by using two different measurement bases: rectilinear and diagonal polarisations. The sender, Alice, selects random bases to encode her bits, and the receiver, Bob, selects random bases to measure each incoming photon. The raw key emerges from Alice and Bob’s bases alignment, which produces matching bits between their measurement results. The key passes through two more stages of error correction and privacy amplification to eliminate all noise and potential eavesdropper information. We establish a fundamental communication model in which Alice transmits n qubits to Bob, who measures them in a basis chosen with probability p. The estimated number of basis matches will reach n⋅p in the average case. Assuming p = 0.5 for BB84, the raw key length k after sifting is given by Eq. ( 1 ). 1 After the sifting phase, error estimation is carried out to compute the Quantum Bit Error Rate (QBER), which is defined as the ratio of mismatched bits to the total sifted key length as in Eq. ( 2 ). 2 where e denotes the number of errors detected during reconciliation. A high QBER indicates the presence of noise or a potential eavesdropper. The key becomes useless when QBER values surpass the established limit of 11% for the BB84 protocol. The implementation of quantum repeaters enables QKD networks to operate over distances beyond the maximum range of photon loss in optical fibres for smart city networks. The devices enable long-distance entanglement distribution via entanglement swapping and quantum memory. In an entanglement-based QKD protocol such as E91, an entangled-photon-pair source distributes one photon to Alice and the other to Bob. The measurement correlations between Alice and Bob enable them to generate a key without needing to trust Alice’s preparation process. QKD system performance in real-world applications relies on three essential factors, which include photon loss, detector efficiency, and channel noise. The secure key rate , which represents the final usable key bits per second, is affected by the transmission efficiency η, the repetition rate r, and QBER as in Eq. ( 3 ). 3 where H(⋅) denotes the binary entropy function. The secure key rate depends on QBER reduction, as urban areas experience signal interference and environmental changes that affect communication stability. Quantum communication provides a novel mechanism for safeguarding vital channels in smart city networks. QKD offers proven security. At the system scale, it encounters several issues, such as stable channels, when communicating with classical networks. These issues are essential to our proposed architecture, as it integrates QKD into a hybrid quantum–classical communication system for intelligent city networks. Federated learning for privacy-preserving threat detection Real-time threat detection systems are needed because smart cities have thousands of connected edge devices and systems. Traditionally, intrusion detection based on centralised ML requires collecting raw data from various sources at a single central server. The method poses significant privacy risks, requires massive data transfers, and introduces compliance challenges across health care, transportation, and law enforcement. In this paper, we propose a Federated Learning (FL) system that preserves the data privacy of intelligent city nodes by jointly training the model without sharing the original data. Local models are developed by individual nodes that first train on their datasets and then communicate model weights or gradient data to a central aggregator on a set schedule. Finally, the aggregator generates a new global model using Federated Averaging algorithms. It is sent in return to each of the contributing nodes. This approach protects sensitive information by maintaining health records, mobility traces, and sensor logs at their sources, while also preserving data privacy in the global security infrastructure. Quantum-secure intelligent city networks benefit from this method because it works best in this particular environment. Innovative city zones train their anomaly detection models using a dual approach that combines classical data (network logs, access attempts, and traffic metadata) with quantum-derived parameters (photon loss rates and QBER fluctuations). The system combines traditional security features with quantum-detection capabilities to identify both standard cyber threats and quantum-specific security breaches, such as channel tampering and side-channel exploitation. The local loss function at node i can be defined as in Eq. ( 4 ). 4 The local loss function at node i is defined by Eq. ( 4 ) which includes is the number of training samples at node i, and is the model parameters w, and ℓ is the loss function ℓ (e.g., cross-entropy). The global model update after aggregation across N nodes is computed as in Eq. ( 5 ). 5 where is the local model at node, and n= is the total number of data samples across all nodes. Differential privacy and secure aggregation in the federated anomaly detection process further improve performance by preventing the sharing of sensitive data patterns during model updates. This method safeguards the transmission of model updates through communication channels secured by quantum cryptography, which employs quantum key distribution (QKD) to shield against interception and tampering. From the perspective of the urban intelligent city node, the federated learning system provides strong, flexible operational capabilities that can cope with a wide range of computational power, networking speeds, and data integrity across urban cities. The system enables participation at different times with partial data, and does so patiently without being dependent on any single system component. A Federated Learning system operates as a scalable, privacy-preserving, real-time threat-detection system, aligned with the need to adapt to changing threat patterns to keep the system resilient. Federated Learning system operates as a scalable and privacy-preserving real-time threat detection system on quantum security-enabled intelligent city networks. This research meets its security and privacy goals with the proposed communication architecture, which facilitates secure collaboration across numerous administrative and infrastructural domains. Table 2 contains all mathematical symbols in the smart city deep learning framework in the paper, along with their descriptions and explanations. Table 2. Mathematical notations and their descriptions used in the proposed framework. Symbol Description QBER Quantum bit error rate, measuring inconsistency in shared key bits Quantum channel health score, computed from QBER, SNR, and photon loss. SNR Signal-to-noise ratio of the quantum or classical communication channel L Photon loss rate or attenuation in the quantum channel α, β, γ Weight coefficients for computing θ Threshold for deciding whether to use QKD or fallback to PQC Encryption selector; 1 for QKD, 0 for PQC D Size of the data payload to be encrypted/transmitted Encryption throughput for method m∈{QKD, PQC} Encryption time for method m Probability of secure transmission success Classical channel reliability (between 0 and 1) Authentication strength or confidence based on Q-PUF Context vector for RL agent at node i: includes trust, latency, etc. Local trust score at node i Quantum link stability estimate at node i Priority score or role-based communication weight at node i Action taken by RL agent at node i Q( , a) Action-value function estimating expected reward for action a Reliability score for neighbor node j Selected next-hop node for forwarding λ Minimum threshold for accepting a neighbor as a next hop Local training loss for FL model at node i Number of samples at node i Model output for input with parameters w Ground truth label for sample j Global model parameters after aggregation at round t Differential privacy noise magnitude at node i Local model update success rate QKD channel integrity during FL communication for node i Probability of successful collaborative detection round C Challenge issued in Q-PUF authentication R Response generated by the device using its Q-PUF Reference response stored during Q-PUF enrolment (⋅) Hamming distance between two binary strings δ Acceptance threshold for Hamming distance during authentication Error rate due to quantum noise during authentication Network delay or packet loss during authentication exchange Trust score of the device participating in authentication Probability of successful device authentication Open in a new tab Materials and methods The general design of the proposed framework is described in this section, which includes architectural components, algorithms, and simulation setup. This research leverages the power of quantum processes to find efficient solutions to smart city scheduling and security problems through a system that integrates quantum-inspired mechanisms, fuzzy-enhanced optimisation, and federated learning modules. This part establishes the datasets, evaluation metrics, and implementation methods used to validate the proposed method. System overview: QSC-Net architecture with MF-MBO-based scheduling and optimization We propose a new type of system, QSC-Net (Quantum Secure Communication Network), comprising both quantum and classical components, establishing a scalable architecture to grant end-to-end secure communication between multiple smart cities operating in tandem. As urban infrastructure increasingly relies on real-time data flows from across multiple sectors — from traffic management and mobility, to healthcare, energy, public safety and IoT citizen services — the need for a reliable, secure and interoperable communications fabric has never been greater. Meeting this need, QSC-Net seamlessly combines quantum key distribution (QKD) and post quantum cryptographic fall back (PQC) with an AI-driven trust-aware routing mechanism, a federated anomaly detection system and a centralized scheduling and optimization module, based on MF-MBO, under a single unified framework. MF-MBO serves as the internal resource and task scheduling engine in the architecture, facilitating efficient workload execution and balanced utilization of distributed virtual machines within the quantum-resilient external security constraints. The overall architecture of the quantum-secure communication system operated in smart cities is shown in Fig. 1 . Fig. 1. Open in a new tab Quantum-secure communication framework for smart cities (QSC-Net architecture). The heart of QSC-Net is a network of QKD nodes from data control centres in every smart city. We define these nodes as trust anchors based on quantum mechanics, performing key generation and distribution through BB84 and E91 protocols. A classical networking interface is co-located with each QKD node, enabling integration with existing urban optical fibre or 5G infrastructure. The quantum layer produces keys and establishes secure connections and the classical channel carries real data. This enables simultaneous operation of the channels, allowing for continuous secure communications in situations where high data transfer rates are required. Quantum repeaters, in addition to entanglement-swapping techniques, provide building blocks to distance entangled photon pairs thus carrying quantum information for secure communication between smart cities at disparate locations. Repeaters work as trusted nodes that do state-refresh operations to send entanglement states from one city to another without loss of quantum information for key generation between non-fibre-optic-connected cities. Based on the stability, QSC-Net employs a backup mechanism which will trigger the post-quantum cryptographic (PQC) algorithms, namely, the lattice-based Kyber encryption, when the quantum channels are affected by decoherence or atmospheric interference. A trust management and routing system, powered by a reinforcement learning (RL) agent, and connected to each QKD node, allows the RL-trained agent to select the most secure and reliable routes for communication over these lines. The action policy of the rate limited (RL) model directly learns from real-time feedback from multiple parameters associated with the network such as quantum link stability, packet loss, trust score of neighbouring nodes, and latency. The system adapts its path to changes in the environment whilst its data gets better protected from start to finish. Architecture supports federation learning based basis of anomaly detection using smart city nodes which computes local intrusion detection model by combining quantum enhanced logs with standard network based metrics. Post aggregation process takes place at fixed temporal intervals over quantum-secured communication channels to output a global model where local data confidentiality is maintained, however able to learn from new patterns of threats. QSC-Net enables device-level authentication through the integration of Quantum Physical Unclonable Functions (Q-PUFs) with its IoT endpoints. They protect the network against impersonation and node forgery, as only tamper-proof legal devices are allowed on the network through hardware-anchored identities. The system design is composed in such a way that it should be extended but scalable in such a manner that in case of any failures it should handle it gracefully. QSC-Net balances efficiency, practicality, and security in the construction of the future smart cities by integrating quantum key distribution (QKD) with post-quantum cryptography (PQC) and artificial-intelligence (AI)-based routing and federated detection systems to offer defense against classical and quantum threats. Secure communication workflow As shown in Fig. 2 , the QSC-Net framework integrates a trusted communication system, which functions as an adaptive pipeline for switching between quantum-notarised and post-quantum encrypted approaches depending on the channel stability and system performance requirements. It fully secures smart city communications through its traditional secure method—the security properties are preserved as it transitions from the security state to the quantum; quantum systems are never stable. The proposed workflow includes quantum key establishment and channel assessment factors, then adaptive encryption and message transmission, with authentication. Fig. 2. Open in a new tab Quantum-enhanced secure communication workflow in smart city IoT. Two quantum key distribution (QKD)-connected smart city gateways demonstrate their functionality via a QKD session. Depending on the available infrastructure for quantum bit transfer, either the BB84 protocol or the E91 entanglement-based scheme will be implemented. Processing of quantum key material involves three separate steps: sifting, followed by error reconciliation and privacy amplification. According to Eq. ( 1 ), the total number of quantum bits sent is used, and only half remain available after sifting. The quantum bits are then processed for QBER calculation, which analyses the noise coming from the system and any eavesdropper. The key becomes useless if the security QBER threshold is breached. It discards keys if the QBER exceeds the security measurement threshold and triggers a different set of security protocols. To determine the viability of continuing with QKD, the system calculates a quantum channel health score , incorporating three critical parameters: QBER, signal-to-noise ratio (SNR), and photon loss L. This is formalized as in Eq. ( 6 ). 6 The values of α, β, and γ need adjustment based on both environmental conditions and device-specific factors. If the computed value is greater than or equal to a predefined threshold θ, QKD is deemed usable. The decision process uses a binary encryption selector which is defined as in Eq. ( 7 ). 7 The system uses a lightweight symmetric encryption algorithm (such as AES-256) with the quantum-generated key when QKD becomes available ( =1). The system switches to post-quantum cryptographic (PQC) encryption through lattice-based Kyber when QKD becomes unstable or unavailable ( =0). The selection method produces different latency results based on the particular approach which researchers use to calculate encryption time for payload D through method m∈{QKD, PQC} as in Eq. ( 8 ). 8 The parameter represents the encryption throughput for the selected method. The encrypted payload gets sent through classical communication networks which include optical fiber, and 5G systems. The packet contains an authentication tag which allows the recipient node to verify the sender through Quantum Physical Unclonable Functions (Q-PUFs). These device-specific quantum fingerprints provide non-reproducible authentication, offering stronger resilience than conventional credential-based systems. The overall secure transmission success probability can be expressed as a function of the classical channel reliability , Q-PUF-based authentication score , and QBER as expressed in Eq. ( 9 ). 9 The system sends back a status acknowledgment after the message has been properly decrypted and validated. The AI-based control layer receives this feedback to modify its routing approach and trust evaluation system for upcoming communication sessions. The system runs multiple times to process each transmission while adjusting its operations to changing quantum and classical network conditions during active operation. The secure communication system achieves quantum-resistant data transmission through its smart city infrastructure by merging real-time channel assessment with hybrid cryptographic backup systems, and AI-based decision control. AI-driven routing strategy The secure and efficient transmission of data between distributed subsystems remains essential for smart cities to function properly in their current dynamic environment. The traditional systems depend on static routing protocols which cannot handle evolving network traffic patterns and changing quantum link performance and node trust levels and external disturbances. The QSC-Net system solves these problems by using reinforcement learning (RL) to develop a routing system which adjusts its routing decisions through immediate communication environment and security situation feedback. The routing strategy based on quantum-aware reinforcement learning appears in Fig. 3 . Fig. 3. Open in a new tab Quantum-aware reinforcement learning-based routing strategy. Each QKD-enabled smart city gateway contains a deep reinforcement learning agent which serves as the fundamental component of this routing system. The agent observes a context vector at node i, composed of multiple features relevant to secure communication. The system uses local trust scores together with quantum link stability estimates , node-level priority metrics , and classical metrics including buffer occupancy, latency, and signal strength. The context vector is formalized as in Eq. ( 10 ). 10 The latency and packet delay variations that occur at node i are represented by and . The RL agent applies a policy function π(a∣ ) to select the optimal action from the routing options A={forward, buffer, drop}. The policy system undergoes training to reach maximum cumulative reward by using a function which combines performance metrics (throughput and delay) with security indicators (quantum channel integrity and trust violations). The routing decision function is expressed as in Eq. ( 11 ). 11 The function Q( ,a) represents the estimated value of performing action a in context . Q-learning or deep Q-network (DQN) algorithms learn this value. The agent selects a next-hop node from its neighbor list based on reliability scores when forwarding packets after choosing an action. The reliability of each neighbor is determined by combining their trust levels with quantum link stability and priority level through weighted calculations as in Eq. ( 12 ). 12 The node with the highest score that meets predefined thresholds is selected for transmission: 13 The reliability threshold λ ensures the next hop node fulfills both trust requirements and stability criteria. The RL agent’s policy updates remain continuously active via a reward feedback loop during execution. The agent is rewarded for successful transmission (signalled by an acknowledgement); in contrast, it is punished with negative rewards for packet loss or delay (due to buffer overflow or denial-of-service attack). As the agent learns, it ends up with an optimal policy that delivers the best performance, system resilience, and security. In such a decentralised routing model, the multi-city dimension of scaling is feasible, as nodes can learn independently while exchanging parameters. To mitigate performance drops in the RL agent due to insufficient training or nonresponsiveness, the system’s backup protocols invoke rule-based routing with fixed trust thresholds. With QSC-Net dynamically adjusting network and quantum link conditions in real time, the AI-based routing system within a smart city communication system can detect untrusted paths or unstable links, leveraging QSC-Net’s dynamic adjustment capabilities. This would build a solid foundation, ensuring the secure urban communication infrastructure remains stable under any circumstances. Federated anomaly detection module The interconnected smart cities generate massive amounts of data from distributed nodes, making centralised threat detection systems unworkable due to latency issues, bandwidth constraints, and privacy concerns. The QSC-Net framework solves these problems by implementing a federated learning (FL)-based anomaly detection system that supports privacy-preserving cybersecurity operations through collaboration between city-level infrastructure networks. Smart city nodes train their anomaly detection models locally through this method by using their own data, which includes quantum channel logs and trust scores, and classical network traffic, without needing to send raw data to a central server. Figure 4 presents federated learning-based anomaly detection in quantum-secure smart city networks. Fig. 4. Open in a new tab Federated learning-based anomaly detection in quantum-secure smart city networks. The smart city nodes operate independently while running a lightweight intrusion detection system (IDS) that uses a hybrid feature set for training. The system operates using standard network parameters (including packet size, frequency, and latency anomalies) and quantum-specific measurements that monitor changes in QBER, photon detection speed, and quantum key exchange issues. Equations ( 9 )–( 13 ) collectively define the adaptive position-updating and hybrid control mechanism, which are unified in the final solution update rule expressed in Eq. ( 14 ). The local model at node i minimises the empirical loss function in Eq. ( 14 ). 14 The number of labeled or pseudo-labeled samples at node i is represented by and is the anomaly detection model which uses weights w and ℓ represents the binary or multiclass loss function based on attack taxonomy. The federated learning system operates through a sequence of rounds which establish its operational framework. Each node sends its model parameters or gradient information to a central federated aggregator through quantum-protected classical channels after completing local training. The global model receives updates through weighted averaging which follows the formula shown in Eq. ( 15 ). 15 The total number of participating nodes is represented by N and n= stands for the complete number of samples. The system guarantees that each node will receive data contributions which match the dimensions of its local data. To enhance robustness and security, the module integrates differential privacy mechanisms that add noise to the local model updates, and secure aggregation protocols that prevent any participant—including the aggregator—from inferring individual model contributions. The model update process employs quantum-generated session keys to maintain synchronization. This defends against parameter poisoning and man-in-the-middle attacks during the update phase. Based on a hybrid architecture consisting of an autoencoder and a shallow classifier or a graph neural network (GNN) that employs logical trust relationships and communication patterns as edges, the anomaly detection model operates independently on all nodes. Anomaly detection is flagged locally, followed by the threat severity score, which updates the global security view using anonymised data aggregation across the entire Network. The model in Eq. ( 16 ) is used to describe the likelihood of completing a secure and successful collaborative detection cycle. 16 The model update success rate is represented by , privacy noise magnitude by and QKD-secured channel reliability for node i is denoted as The system maintains high even when nodes join or leave the network and when update data contains errors. The federated anomaly detection module in QSC-Net provides smart cities with a robust, scalable security system that operates through federated detection. The system operates without centralised data storage while learning from local threats and providing collective learning capabilities through quantum-security-protected privacy methods. The system needs this ability to detect malicious activities before they occur, identify insider threats, and protect against classical and quantum attacks on smart city infrastructure. Quantum-enabled authentication mechanism The fast-changing, decentralised smart city systems depend on device authentication as their primary foundation to protect data security and maintain network stability. The current authentication systems, which use passwords, digital certificates, and biometric identifiers, are vulnerable to forgery and quantum attacks and fail to support large-scale deployment of the Internet of Things (IoT) networks. The QSC-Net system addresses these issues by using its Quantum Physical Unclonable Function (Q-PUF) authentication module to provide secure hardware-based identity verification for all network nodes and edge devices. Q-PUFs are physical unclonable functions that generate unique response patterns from input challenges by combining quantum device manufacturing variations and quantum-level random behaviour in quantum-scale devices. The quantum aspects of phase noise, together with energy-level fluctuations and entanglement-based state differences, produce conditions that no adversary can replicate for practical applications. The integration of Q-PUF technology into devices generates distinctive quantum fingerprints that produce unpredictable responses that cannot be duplicated, making them ideal for secure authentication systems. The QSC-Net system employs a challenge–response protocol for mutual authentication. The control node sends a random challenge C to the device D when it tries to access the network. The Q-PUF embedded in D addresses this challenge by leveraging its intrinsic quantum variations to generate a response R, which is compared against a pre-registered database of valid challenge–response pairs—the authentication process results in a response as in Eq. ( 17 ). 17 and the Hamming distance between R and the stored response is below an acceptance threshold δ as in Eq. ( 18 ). 18 Here, denotes the Hamming distance. The threshold δ is empirically determined based on device stability and environmental factors, ensuring tolerance to minor fluctuations without compromising security. The authentication process in QSC-Net uses quantum session key establishment via QKD to prevent impersonation, replay, and man-in-the-middle attacks. The system uses a new QKD-derived symmetric key to encrypt all challenge-response exchanges and signalling, which ensures both confidentiality and forward secrecy. The authentication system uses time-limited nonce values and single-use challenges to defend against attackers who try to reuse old responses or attack during specific time frames. The authentication success probability for a legitimate device is modeled as in Eq. ( 19 ). 19 The equation includes three parameters which represent quantum noise error rate , authentication exchange network delay or packet loss rate , and device trust score ∈[0,1]. The proposed system shows that authentication robustness depends on the performance of quantum hardware systems and the quality of communication channels. The authenticated devices will receive quantum session tokens from the system, and these tokens will permit the devices to perform specific tasks, e.g., data transfer, routing, and model dissemination, during their lifetimes. Ninety-nine, unlike the language model, can generate new tokens at regular intervals, and a load balancer prevents attackers from using stolen credentials indefinitely. The trust management engine detects devices that fail repeated authentication checks or any suspicious behaviour pattern. This places them on a quantum threat watchlist for further investigation. QSC-Net uses quantum-enabled authentication as part of its communication system to prevent unauthorised devices from tapping into sensitive data exchanges. The role of the model rooted in hardware trust forms the basis of a unique security system for smart city applications that process tons of real-time data — and only with the advent of the post-quantum era will digital certificates be susceptible to attacks. Mathematical perspective of the MF-MBO framework In this section, we provide the formal mathematical underpinnings of our proposed MF-MBO framework to ensure clarity, precision, and reproducibility. This framework comprises fuzzy dominance ranking, quantum-inspired tunnelling to balance exploration and exploitation, and greedy migration for local refinement, and it can handle multi-objective task scheduling. The phrase quantum-inspired, as used in this paper, means classical algorithms inspired by qualitative features (e.g., tunnelling-like escape from local optima via probabilistic acceptance) of quantum computation, but does not involve any quantum hardware or physical simulation of quantum states. Note that the MF-MBO optimisation layer is purely classical. The QKD layer in QSC-Net differs from ours, and thus it has been modelled using conventional BB84/E91 simulation parameters (QBER, photon loss, repeater intervals, etc.) to capture the physical properties of the quantum channel. Consequently, we refrain from using the term quantum computation, limiting ourselves to the terms quantum-inspired (classical) optimisation and simulated quantum communication. Let a candidate solution be denoted as , representing a mapping of tasks to virtual machines. The optimization objective is defined as a normalized multi-objective vector as in Eq. ( 20 ). 20 where , , and correspond to execution time, energy consumption, and load imbalance, respectively. Fuzzy dominance modeling For each objective , a fuzzy membership function is defined as in Eq. ( 21 ). 21 where and are population-wise bounds and is a small constant. The fuzzy dominance degree between two solutions and is computed as in Eq. ( 22 ) 22 where is a tolerance-based comparison function. A solution is considered dominant if , enabling robustness against uncertainty and objective conflicts. Quantum-inspired tunneling mechanism To escape local optima, MF-MBO adopts a quantum-inspired tunneling acceptance strategy. Given a parent solution and a candidate solution , the fitness difference is as in Eq. ( 23 ). 23 If , the candidate is accepted; otherwise, it is accepted with probability as in Eq. ( 24 ). 24 where is a self-adaptive tunneling parameter that decays over generations, ensuring a gradual transition from exploration to exploitation. Greedy migration refinement Greedy migration performs local optimization by redistributing tasks from heavily loaded to lightly loaded virtual machines. Let and denote the most and least loaded machines, respectively. A migration is accepted if the condition is given in Eq. ( 25 ) is set. 25 where is a minimum improvement threshold. This ensures monotonic local improvement while keeping computational overhead bounded. Algorithmic implementation We now introduce the key algorithms that drive the QSC-Net framework. The system is described in three sections – fuzzy-enhanced quantum optimization (to schedule quantum jobs), reinforcement learning (to route jobs adaptively), and federated anomaly detection (with differential privacy). To maintain the clarity and reproducibility, the implementation of integrated smart city communication and security systems relies on logical and mathematical formulations to be followed step-by-step. Algorithm 1 starts a secure communication link between two smart city nodes through Quantum Key Distribution (QKD) while switching to Post-Quantum Cryptography (PQC) when the quantum channel performance decreases. The system allows only authenticated devices that have been verified through Quantum Physical Unclonable Functions (Q-PUFs) to join the secure session. Algorithm 1. Open in a new tab Secure session initialization with quantum channel. The first step requires establishing a QKD session through either the BB84 prepare-and-measure protocol or the entanglement-based E91 approach. The system keeps only the bits which both sender and receiver selected the same basis after photon transmission. The system determines the Quantum Bit Error Rate (QBER) through bit comparison between two parties over a classical communication channel. The key becomes invalid for use when the QBER reaches a certain threshold because it suggests either eavesdropping activities or excessive noise in the system. The quantum channel health score determines channel usability through measurements of QBER together with signal-to-noise ratio (SNR), and photon loss. The system compares this score to the predefined reliability threshold θ. The quantum channel reaches acceptable performance level when the score is ≥ θ. So the QKD key generation process produces a usable key. The algorithm switches to PQC for session key generation through lattice-based cryptosystem Kyber when the condition does not meet the requirement. The initiating node will send a random challenge to the receiving node as soon as the encryption mode selection process is complete. The receiver uses its built-in Q-PUF to analyze this challenge, which results in a distinctive response. The system verifies the response through Hamming distance by matching it with a pre-registered value to confirm hardware-based identity authentication. The system finishes the session by applying the pre-established key together with the selected encryption mode after successful authentication. The hybrid algorithm establishes secure communication through adaptive post-quantum protection by verifying device trust using quantum hardware fingerprints during unstable quantum link operations. As in Algorithm 2 , it operates through a reinforcement learning (RL) based routing algorithm which dynamically adjusts its operations to improve data protection and network stability in quantum-secured smart city environments. The method operates differently from static routing protocols because it enables network nodes to learn from their environment and modify their packet forwarding operations through trust evaluation and active network monitoring. Algorithm 2. Open in a new tab RL-based trust-aware secure routing. Each smart city node is equipped with an RL agent that monitors the local environment and derives a context vector . The vector contains essential routing information, which includes the node’s trust level at present , quantum link stability assessment , priority level , and classical network performance indicators of average latency and delay variation . The three elements serve to evaluate the communication environment which exists at node i. The RL agent employs a policy π(a∣ ) which it has learned to pick the best routing action from its discrete action set {forward, buffer, drop}. The agent uses a Q-function or deep Q-network (DQN) to calculate the expected reward for each action. The chosen action seeks to improve both packet delivery performance and security standards. The agent will discard the packet when it chooses to drop it. The packet remains in buffer storage until the system finds an improved routing route. The agent evaluates the set of neighboring nodes when the action is set to forward. The system determines a reliability score for each neighbor j by combining trust, quantum stability, and priority elements through a weighted sum formula: =α⋅ +β⋅ +γ⋅ . The agent will pick the next-hop node which shows the top reliability score that exceeds the set threshold λ to forward packets through nodes which maintain both security and stability. After forwarding or buffering, the node gets feedback about the result which could be delivery success or a timeout condition. The agent uses this feedback to improve its policy by increasing the strength of successful actions while decreasing the strength of unsuccessful decisions. This allows it to learn continuously. The routing model reaches an optimal solution through its operation because it learns to adjust its routing decisions based on changing quantum channel conditions and trust relationships and network traffic levels. The AI-based secure routing system operates as a real-time intelligent decision system which protects the QSC-Net framework. The system protects against compromised or unreliable nodes while maintaining trust and security, and protecting against routing attacks and quantum link instability. The distributed smart city nodes within Algorithm 3 establish a global anomaly detection model through collaborative training which maintains data privacy and secure communication. The proposed system exploits federated learning (FL) which demands nodes to train models locally on their data before sending encrypted updates to a central aggregator. The entire process is secured using quantum key distribution (QKD) to prevent tampering or interception. Algorithm 3. Open in a new tab Federated anomaly detection with secure aggregation. Every node i that participates in the system starts with the global model weights along with its local dataset , which contains both traditional network traffic data and quantum-based signals, including QBER changes, photon loss behavior, and key negotiation irregularities. The node uses this information to conduct local training by reducing the empirical loss function on its dataset through an autoencoder-based anomaly detection model or a shallow classifier. The local training process produces new model parameters as its output. Each node generates random differential privacy noise with the distribution N(0, \epsilon), which serves as a privacy-preserving mechanism before sending its local model update. The mechanism prevents inference attacks from happening while protecting sensitive information even when someone intercepts the model update. The communication channel achieves information-theoretic security through QKD-derived symmetric key encryption, which protects each model update. The aggregator receives encrypted model updates from all nodes before decrypting them to perform secure aggregation through a weighted averaging approach. The global model is computed based on the contribution of each node, proportional to its data size . The system prevents bias toward small or untrained nodes while maintaining statistical convergence. The updated global model is then redistributed to all participants via the same quantum-secured channel. The nodes start using their final model for real-time anomaly detection after completing T fixed aggregation rounds. The system evaluates detected anomalies at the local level to determine whether they need to activate threat alerts for DDoS attacks, unauthorized access, and quantum-layer disruptions. The method for successful collaborative detection depends on update success rate combined with channel integrity and privacy noise to achieve both accuracy and security. The algorithm provides improved cybersecurity for interconnected smart cities through its scalable quantum-resistant intrusion detection system, which preserves privacy. It operates without central data collection or local system interference. The complete Unified MF-MBO framework incorporating fuzzy dominance, quantum inspired tunneling and greedy migration for the MO task scheduling is shown in Algorithm 4 . Finally, the algorithm performs population initialization and objective normalization, then fuzzy dominance–based ranking to identify good solutions in an uncertain environment. Based on model-based optimization (MBO) inspired readers, the migration operators generate new candidate solutions, while the quantum-inspired tunneling mechanism accepts inferior solutions with a certain probability to escape from local optima and thus improve the exploration. It is followed by a greedy migration phase which is bounded and it refines the selected solutions by locally optimizing load balance and energy consumption. Using iterative elitist selection and self-adapting parameters, schedules continuously optimize toward higher quality and with greater scalability, robustness and reproducibility. Algorithm 4. Open in a new tab Unified MF-MBO framework (Fuzzy Dominance + Quantum Tunneling + Greedy Migration). Although fuzzy dominance, quantum-inspired acceptance and greedy local search have been studied independently in the literature, the novelty of MF-MBO is the special coupled mechanism and coupling strategy for multi-objective scheduling. MF-MBO first presents a tolerance-aware fuzzy dominance degree that stabilizes dominance score computation between conflicting multi-objectives, used consistently and in a stable manner across (i) elite selection, (ii) survivor selection, and (iii) acceptance gating to achieve robustness towards stochastic workloads. Second, the framework combines this dominance score with a self-adaptive tunneling acceptance rule, which turns dominance-based fitness gaps into a mechanism to adaptively modulate exploration intensity between generations, thus creating a controlled escape from stagnation, as opposed to ad-hoc random re-starts. Third, MF-MBO merges a bounded greedy migration heuristic that focuses on the most/least loaded resources based upon a minimum-improvement threshold that guarantees monotonic local improvement while maintaining expected overhead—a key requirement for real-time scheduling. Finally, this combo is not an abstract hybrid: we realize it as one reproducible pipeline (Algorithm 4 ) with indepth complexity/overhead and statistical significance analyses showing that the joint process leads to provable improvements over individual components. Baseline algorithms and fair parameter settings To ensure a fair comparison, all metaheuristics were evaluated under an identical computational budget and stopping criteria. Specifically, GA, PSO, standard MBO, and the proposed MF-MBO were run with the same population size , maximum iterations , and the same objective functions and constraint handling. Each algorithm was executed for 10 independent trials using different random seeds, and results were reported as mean ± standard deviation. For baseline methods, commonly adopted parameter values from prior scheduling literature were selected and kept fixed across all experiments. For MF-MBO, the fuzzy dominance thresholds and tunneling schedule were set once and applied consistently across all workloads. This protocol ensures that improvements are attributable to algorithmic design rather than unequal tuning effort. The parameter settings for all competing baseline algorithms and the proposed MF-MBO framework are summarised in Table 3 . Population sizes, control parameters, and stopping criteria are reported as they are applied across experiments. The table establishes a fair and reproducible comparison, where no performance gain arises from unequal parameter tuning, by enforcing the same evaluation budgets and trial protocols. Table 3. Parameter settings used for baselines and MF-MBO (fair comparison protocol). Algorithm Key parameters (used in all experiments) Fairness constraint GA Population =50; crossover =0.9; mutation =0.05; tournament size = 3; elitism = 2 Same , iterations , objective functions, trials PSO Swarm =50; inertia =0.7; cognitive =1.5; social =1.5; velocity clamp = ± 0.2 Same , , and stopping rules MBO (baseline) Population =50; migration period = 2; partition ratio = 0.5; mutation rate = 0.05 Same evaluation budget and constraints MF-MBO (proposed) =50; =200; fuzzy thresholds ; tunneling ; greedy steps , Same , , trials; single fixed setting across scenarios Open in a new tab Experimental results To demonstrate the performance and security analysis of QSC-Net at both the quantum and classical network layers, extensive simulations were performed to validate the proposed framework. The details are listed numerically, showing how QSC-Net excelled in secure session establishment, adaptive routing, federated anomaly detection, Q-PUF authentication, and dynamic smart city resilience compared to the old baseline methods. Evaluation setup and testbed overview (revised for reproducibility) To validate the practical effectiveness of the proposed QSC-Net framework, a hybrid simulation environment was built that integrates the quantum communication layer with the classical network emulation layer to provide realistic smart city conditions. In the testbed ASQNET, two interconnected smart-city communication infrastructures controlled five city service aspects (traffic management, healthcare, power grids and surveillance) with nodes embedded with four essential components, including quantum key distribution (QKD) modules, an authentication unit based on Q-PUF, a federated anomaly detection client and trust-aware routing RL agents. All experiments were carried out on an isolated host machine to enable the repetition of runtime and communication measurements. The host specification provides for the CPU ID [ 14 ], core count, RAM size, storage type, OS, and implementation stack (Python/NS-3 build). In case virtualisation was needed to emulate heterogeneous edge/cloud nodes, a VM-based testbed was used, with the VM count and resource (vCPU, RAM, disk, and network) profiles fixed. To avoid any configuration-caused variation, the virtualisation platform (KVM/VirtualBox/VMware), VM OS image, and networking mode (virtual switch/bridged) were kept the same across trials. Simulated BB84-type QKD for intra-city key generation and E91 entanglement-based protocols for inter-city quantum communication were deployed at the quantum layer. Different quantum parameters were adjusted to mimic environmental factors and realistic link conditions. The evaluation specifically ranged the quantum bit error rate (QBER) from 1% to 12%, and simulated photon loss across fibre pathways of 20 km to 100 km. It included channel noise caused by atmospheric disturbances and photon-detection faults. To enable entanglement swapping and long-distance scalability, quantum repeaters were placed every 40 km. A classical layer is used in NS-3 to simulate high-capacity, secure communication over 5 G and optical fibre channels. Three traffic load scenarios were evaluated: Low (10 packets/sec), Medium (100 packets/sec), and High (500 packets/sec), representing streaming video, sensor telemetry, alert messages, and encryption/command-and-control data. We explicitly controlled for workload sizes by varying (i) node count, (ii) offered load, (iii) message sizes per traffic type, and (iv) simulation duration. To mimic stochastic city traffic patterns, urban message arrivals were modelled as Poisson processes. Each simulation run took 1,000 s, and the experiment was repeated 10 times with different random seeds to estimate variance (mean ± standard deviation, where appropriate). Reproducibility: The entire simulation configuration, platform/testbed details, and workload sizes are summarised in Table 4 . Table 4. Simulation configuration, testbed specifications, and workload parameters. Component Configuration/value Host hardware (for reproducibility) CPU: Intel Core i7-12700 K; Cores/Threads: 12 C/20T; RAM: 32 GB; Storage: 1 TB NVMe SSD Host software OS: Ubuntu 22.04 LTS (64-bit); NS-3 version: NS-3.38; Language/runtime: C++ (GCC 11.4) + Python 3.10 Virtualisation platform (if used) Platform: KVM/QEMU; VM OS: Ubuntu 20.04 LTS (64-bit); VM networking: Bridged virtual switch (Open vSwitch) VM configuration (if used) No. of VMs: 10; vCPU per VM: 2 vCPU; RAM per VM: 4 GB; Disk per VM: 40 GB (dynamic) Number of smart city nodes 5 (service domains: traffic, healthcare, grid, surveillance, control) QKD protocols BB84 (intra-city), E91 (inter-city with entanglement swapping) QBER range 1% – 12% Photon loss range Simulated over 20–100 km Quantum Repeater interval Every 40 km Classical simulator NS-3 Traffic types Video streaming, telemetry, alerts, control messages Traffic loads Low (10 pkt/sec), Moderate (100 pkt/sec), High (500 pkt/sec) Message sizes (workload size) Video: 1200 bytes; Telemetry: 128 bytes; Alerts: 256 bytes; Control: 512 bytes Arrival model Poisson arrivals (rate set by load level) Simulation duration/repetitions 1,000 s × 10 trials (distinct random seeds) RL agent model Deep Q-Network (DQN), episodic learning RL features Trust, quantum link health, latency, buffer state Federated learning model Autoencoder-based anomaly detection FL communication QKD-secured; differential privacy ( ) Authentication method Q-PUF with Hamming distance threshold ( ) Open in a new tab Our RL-based routing agents employed DQNs to dynamically select optimal next-hop nodes for routing packets over multiple paths in response to changing network conditions, represented in a multidimensional context vector containing information such as trust scores, quantum link health (as determined by our implemented quantum protocol), latency, and buffer occupancy for each node. The training approach was episodic, using reward signals indicating successful delivery, minimising latency, and the rationale for avoiding compromised nodes. In the federated learning process, each node first trained a local anomaly detector on hybrid feature sets comprising quantum logs (e.g., QBER anomalies and entanglement mismatch indicators) and classical traffic indicators (e.g., packet timing irregularities and port-scan patterns)—secure aggregation: differential privacy to hide model updates, combined with QKD-secured channels. Q-PUF challenge–response pairs (CRPs) were used for authentication, and verification was performed using Hamming distance thresholding. The quantum noise levels were tuned to mimic distortion at the physical and transmission layers, enabling evaluation of complete session establishment, intelligent routing, privacy-preserving detection, and device authentication under various operating conditions. Secure session establishment and encryption performance The QSC-Net framework demonstrates its ability to adapt its encryption operations in real time based on the current state of the quantum communication channel. The evaluation of the QSC-Net framework’s adaptive encryption mechanism performance used BB84-based QKD and Kyber-based PQC together. The QSC-Net framework’s evaluation of its real-time encryption method used a test combining BB84-based QKD with Kyber-based PQC. Each session initialisation began with photon transmission over the quantum channel, followed by QBER estimation. The system dynamically selected between quantum-derived symmetric encryption or PQC fallback based on the threshold . Under 4 dB/km photon loss and 40 km distances QKD systems establish keys with less than 5% QBER. The system reverts to Kyber encryption when QBER exceeds 11% over 100 km of transmission or in environments with high loss rates. Table 5 presents the success rates for QKD-based session establishment across different distance ranges. Table 5. Secure session establishment metrics across encryption modes. Metric QKD mode (AES-256 with quantum key) PQC mode (Kyber-1024) Avg. QBER at 20 km 2.7% Not Applicable Avg. QBER at 80 km 10.6% Not Applicable Session establishment success rate 91.3% (at ≤ 60 km) 100% (fallback) Avg. key generation time 0.83 s 1.42 s Payload encryption time (2 KB) 0.47 ms 1.03 ms Avg. authentication success rate (Q-PUF) 98.6% 98.6% False rejection rate (FRR) 1.1% 1.1% Open in a new tab Encryption latency was measured as the time taken to encrypt a 2 KB payload under each mode. With pre-computed AES-256 keys, symmetric encryption (based on quantum principles) operates much faster, whereas Kyber-based key encapsulation and decapsulation is much slower. However, this mechanism, which falls back, maintains secure communication without session interruptions in the system. During session provisioning, device authentication is performed at the system level using the Q-PUF technology. It yields an authentication success rate of 98.6% across 1000 normal and noisy test runs, with false rejection rates below 1.1%. The findings demonstrate the robustness of hardware-based identity verification systems against ordinary quantum interference. Comparative metrics for QKD and Kyber encryption modes are displayed in Table 5 , including their success rates, latency, and authentication strengths. The results demonstrate that QSC-Net ideally enables quantum-secured channels wherever possible, but reverts to post-quantum cryptography in less optimal settings. Stable authentication success rates across modes indicate the system’s reliable, modular design. Figure 5 shows two visualisations that demonstrate how quantum and post-quantum encryption methods operate in the QSC-Net. QKD AES-256 vs. Kyber-1024 (PQC): As the first subfigure (a) depicts, the performance of QKD AES-256 and Kyber-1024 (one of PQC modes) in both encryption and key generation speeds. Using symmetric keys that require very little processing time to encrypt payload data, the encryption latency is only 0.47 ms. In comparison, the key generation time was 830 ms, including photon transmission, QBER estimation, and reconciliation. In terms of processing requirements, the key encapsulation and decapsulation operations in Kyber-1024 PQC mode took about 1.03 ms and 1420 ms for encryption latency and key generation, respectively, while QKD systems exceeded the encryption threshold (i.e., 500 s) of conventional methods. Environmental conditions degrade the key exchange process. Fig. 5. Open in a new tab Performance comparison of encryption modes in QSC-Net framework. The success rate of session establishment is illustrated in subfigure (b) as a function of the transmission distance for both encryption modes. As expected, the QKD-based session success rate decreases monotonically with distance, from 99.1% at 20 km to 50.2% at 100 km, due to higher photon loss and QBER in longer or noisier channels. Error-correction algorithms break down, making the QKD system inoperative beyond 60 km. Except, the Kyber-based fallback works flawlessly every time; it is agnostic to the network environment and performs well during quantum link degradation. This way, the system achieves lower latency, but requires more computing and network resources. These visualisations show that QSC-Net bridges performance and security: QKD provides fast encryption in ideal conditions, while PQC provides security in adverse conditions. QSC-Net maintains dynamic coterminous operation to automatically scale and provide permanently connected, secure communication across intelligent city networks. Routing efficiency and adaptive security gains To assess the performance of the RL-based secure routing mechanism in QSC-Net, we conducted multiple training episodes under dynamic network and trust conditions. According to the classification, the performance assessment was conducted for AODV and the new protocol, which provides security services for routing. Table 6 shows the performance of each routing system. As for packet delivery ratio (PDR), QSC-Net outperformed AODV by 94.6% compared to 86.3%. Because the network agent selects routes that avoid congested or unstable network links, the average transmission latency from QSC-Net is 48.2 ms, compared with AODV’s 61.7 ms; thus, QSC-Net achieves lower latency. Table 6. Routing performance comparison between QSC-Net and AODV. Metric QSC-Net (RL-based) AODV (baseline) Packet delivery ratio (PDR, %) 94.6 86.3 Average latency (ms) 48.2 61.7 Routing convergence time (sec) 12.5 Not Applicable Next-hop trust compliance (%) 98.1 72.4 Open in a new tab We trained the RL agent at each smart city node over 20 episodes, using the state features of trust scores, quantum link stability, and network latency. The learning curve in Fig. 6 depicts an agent gradually improving its performance by learning a routing policy. While the AODV baseline system utility level remained constant across all episodes, the RL-based model achieved higher utility by converging to a near-optimal policy after 15 episodes. It demonstrates the system’s capability to identify protected network paths that ensure operational viability by simulating live network traffic. Fig. 6. Open in a new tab Routing efficiency and adaptive security gains in QSC-Net compared to AODV. On the next-hop level, the computed routing system of QSC-Net obeyed a compliance (the minimal trust level a routing decision is expected) of 98.1%, meaning almost all default routing decisions conform to the trust required at a next-hop level. Because of a lack of integrated trust mechanisms, routing paths in AODV fail to satisfy security restrictions at a rate of 72.4%. QSC-Net obtains a stabler routing policy after 12.5 s of training, while AODV cannot learn or converge. QSC-Net: an Approach for Quantum-Enabled Intelligent City Networks with Optimized Routing and Multiple Trust Level Based Communication Security Adaptation for Enhanced Network Performance. Anomaly detection accuracy and privacy-preserving behaviour The QSC-Net framework assesses the federated anomaly detection system to detect cyber-physical attacks in inter-system communication within intelligent city communication networks. Four primary threat types were analysed: DDoS, spoofing of MAC and IP addresses, quantum link, and unauthorised insider access to edge gateways and local nodes. The architecture used federated learning, with device-side local model training and encrypted parameter updates sent to a central aggregator. We tested two configurations: one without additional noise, providing only QKD-secured aggregation (standard FL), and one that added differential privacy by applying Gaussian noise to the update before sending it (DP-FL). Hence, QKD was the main technology for defending against interception attacks on communication systems, and differential privacy was the technology for protecting local data from reconstruction or the extraction of sensitive information. Results were analysed using conventional classification metrics (precision, recall, and F1-score) for each of the four attack types. The detection performance results are shown in Table 7 , where both configurations are compared. Apart from that, introducing differential privacy results in a slight reduction in classification metrics, but it is a good trade-off since we gain more privacy. These model updates were then sent via secure, tamper-resistant QKD transmission and protected against central data aggregation via federated learning. Table 7. Performance of federated anomaly detection module with and without differential privacy. Threat type Precision (No DP) Recall (No DP) F1-score (No DP) Precision (DP) Recall (DP) F1-score (DP) DDoS attack 0.942 0.913 0.927 0.906 0.887 0.896 Spoofing 0.921 0.889 0.904 0.884 0.859 0.871 Quantum link abuse 0.967 0.944 0.955 0.933 0.917 0.925 Insider misbehavior 0.936 0.911 0.923 0.899 0.878 0.888 Average 0.9415 0.91425 0.92725 0.9055 0.88525 0.895 Open in a new tab Across all binary classification tasks, the system achieved relatively high accuracy, representing one of the most successful results for quantum link abuse detection. By applying privacy-preserving methods to the first two layers, our detection system can still operate in real-world smart city systems, maintaining functionality while remaining privacy-preserving. We present the performance of our federated anomaly detection system for detecting four threat categories—namely, DDoS Attack, Spoofing, Quantum Link Abuse, and Insider Misbehaviour—in both DP and non-DP setups, expressed as percentages, as shown in Fig. 7 . Fig. 7. Open in a new tab Anomaly detection performance metrics with and without differential privacy in federated learning for QSC-Net. The precision metric, therefore, indicates how well the model performed at predicting real threats. We show in our analysis that the system achieves very high precision for all threats where DP is not deployed (e.g., > 0.967 for Quantum Link Abuse and 0.942 for DDoS). This means that instead of relying on only one individual’s data, DP introduces noise into predictions or analyses, which causes a degradation in precision when DP is implemented, in this case, from 0.933 to 0.906, due to the extra noise introduced by the privacy protection process. Recall — the ability to accurately identify all real threats. Without DP, the system produced recall scores ranging from 0.889 to 0.944. Implementing DP leads to a small and sustained (up to 3–4%) decrease in the detection sensitivity for all threats due to the perturbations induced for privacy preservation. Although DP also works well for the F1-score, it is a good combination of precision and recall values. For the Quantum Link Abuse, the F1-score drops from 0.955 (No DP) to 0.925 (DP), which is still a slight improvement. For example, Insider Misbehaviour falls from 0.923 to 0.888. Overall, the federated model provides stable and homogeneous performance in detecting all threat classes, although approaches based on differential privacy result in performance degradation. The integration of quantum key distribution (QKD) with federated learning (FL) systems establishes secure communication channels that protect against unauthorised model manipulation and information leakage during training. The experimental findings show that the QSC-Net anomaly detection system maintains high performance levels despite the known privacy-reducing effect of differential privacy. The research demonstrates that federated anomaly detection systems function effectively when combined with quantum-protected, secure channels in smart city infrastructure. The acceptable differential privacy trade-off maintains F1-scores above 88% for all cases. The QSC-Net anomaly detection system delivers both data privacy and excellent detection performance for real-world secure distributed intelligent city networks. Authentication robustness and Q-PUF validation The quantum-physically unclonable function (Q-PUF) based authentication system within QSC-Net underwent testing for accuracy and false rejection rate (FRR), and its performance against quantum noise and Hamming distance thresholds. The system makes authentication decisions by measuring the similarity between challenge–response pairs through a normalised Hamming distance metric, which follows the inequality as in Eq. ( 26 ). 26 where the predefined threshold, and , represent the original and received Q-PUF responses respectively. Additionally, noise in quantum channels is modeled by quantum error probability , simulating real-world decoherence and photon loss. The results, as summarized in Table 8 , demonstrate that Q-PUF authentication achieves an accuracy of 98.4% under low noise ( ) and a reasonable threshold . Even under increased noise ( ), the system maintains over 94% accuracy. FRR remains below 5% in most scenarios, confirming the system’s reliability. In contrast, RSA-based identity schemes degrade significantly in adversarial and noisy settings, with accuracy dropping to ~ 84.2% and FRR exceeding 12%. Table 8. Q-PUF vs. RSA authentication performance under quantum noise and varying hamming thresholds. Scheme Quantum Noise Hamming threshold Accuracy (%) FRR (%) Q-PUF 0.02 0.10 98.4 1.6 Q-PUF 0.05 0.10 96.1 3.2 Q-PUF 0.07 0.12 94.3 4.7 RSA-based ID 0.00 N/A 91.5 7.6 RSA-based ID 0.05 N/A 84.2 12.3 Open in a new tab This unique entity Q-PUF system which is based on quantum resistance and physical uniqueness can be a very powerful defense system against real world attacks in quantum communication systems. The system has outperformed compared to existing body shape based behavioural characteristic analysis methods by achieving very low false rejection rates and preserving high authenticating accuracy, hence it can easily be used for decentralised smart city authentication systems. Q-PUF provides enhanced security capabilities, adaptable operations, and forward secrecy compared to conventional RSA systems, which are susceptible to quantum attacks due to their inability to mitigate quantum noise. In Fig. 8 , we compare the authentication strength of the proposed Q-PUF (quantum physical unclonable function) with that of the classical RSA-based identity schemes across different values of the quantum noise rate (εq), using Eq. ( 26 ) to evaluate Q-PUF performance. Fig. 8. Open in a new tab Robustness of Q-PUF authentication under quantum noise compared to RSA-based identity schemes. From subfigure (a), we can see that even with the increasing quantum noise, our Q-PUF method still shows high authentication accuracy. Q-PUF gains a 98.4% accuracy at εq = 0.02, and the accuracy slowly reduces to 96.1% and 94.3% at εq = 0.05 and εq = 0.07, respectively. On the other hand, RSA-based schemes experience both sharper declines in accuracy from 91.5% to 84.2% with increasing noise, which shows that they are more sensitive to signal degradation in quantum communication layers. (a) and (b): FRR shown over the same range of noise, FRR grow in small numbers, from 1.6% to 4.7% with increasing εq, but Q-PUF shows high stability and low FRR – making it more reliable. In contrast, RSA-based approaches show a much sharper increase in FRR, with FRR at εq = 0.05 rising to 12.3%. The unpredictability of channel noise also highlights the quantum-resilient traits of Q-PUF in smart city networks. Q-PUF has been demonstrated to provide better authentication performance and is a strong defence against quantum-based operational attacks. This mechanism must maintain its operational condition so that, during the quantum relay of identity verification in smart cities through error-bounded channels, it can withstand environmental disturbances. Compared to standard RSA-based identity systems, the Q-PUF system provides higher levels of protection and greater immunity to signal interference. Experimental data show that quantum-enhanced authentication systems work well in new secure communication systems. Comparative baseline analysis A comparative analysis was completed to determine QSC-Net performance compared to three well-known smart city communication systems: an SDN-IDS, a Conventional BB84-based QKD Overlay Architecture, and a Secure IoT Gateway with Centralised Trust Model. The baselines are outdated approaches that rely on outdated secure communication, trust establishment, and anomaly detection systems. QSC-Net outperformed all evaluation metrics across transmission security, trust setup latency, adversarial responsiveness and throughput under attack, and node scalability. The QSC-Net system achieved a secure packet delivery rate of 96.8%, which was better than SDN-IDS (88.1%), the BB84-QKD overlay (90.7%), and the secure IoT gateway model (85.3%). QSC-Net achieves better performance through its single-system design, which combines intelligent RL routing with QKD and distributed anomaly detection. Table 9 presents the evaluation results of QSC-Net against SDN-IDS, BB84-QKD, and the Secure IoT Gateway for security performance, latency, and scalability. Table 9. Comparative performance of QSC-Net vs. classical baselines. Metric QSC-Net SDN-IDS framework BB84-QKD overlay Secure IoT gateway Secure packet delivery (%) 96.8 88.1 90.7 85.3 Device-level trust establishment time (ms) 22.3 45.6 41.8 49.2 Attack response latency (ms) 11.2 23.7 19.8 27.1 Throughput under attack (Mbps) 78.6 61.4 65.9 56.3 Nodes supported (scalability test) 500+ 300 350 280 Open in a new tab With QSC-Net, the delay for establishing trust was only 22.3 ms with Q-PUFs, which was almost 50% lower than the centralised IoT gateway (49.2 ms), SDN-IDS (45.6 ms), and BB84 overlay (41.8 ms). The QSC-Net system responds within 11.2 ms of an attack, enabling it to detect ongoing attacks quickly. The response times of the BB84 overlay and SDN–IDS systems were 19.8 ms and 23.7 ms, respectively. Throughput rates were measured as high as 78.6 Mbps during adversarial testing for the QSC-Net system, outperforming SDN-IDS (61.4 Mbps), BB84 (65.9 Mbps), and the centralised gateway (56.3 Mbps) (Li et al., 2020). Although QSC-Net maintained continuous stability with 500 active smart devices during the scalability test, the SDN-IDS, BB84, and the centralised model reached their upper limits of 300, 350, and 280 nodes, respectively. The outcome proves that QSC-Net offers an optimal architecture for full quantum security transmission, adaptable trust networks, and high network performance for future intelligent city networks. The performance comparison of QSC-Net with the three baseline systems (SDN-IDS, BB84-QKD, and a Secure IoT Gateway with RSA-based authentication) is presented in detail in Fig. 9 . We investigate secure packet delivery, the time required for a node to establish trust, the time to respond to an attack, throughput during attacks, and system scalability, using these as the five principal dimensions to evaluate our proposal. Fig. 9. Open in a new tab Comparative performance analysis of QSC-Net with existing secure communication frameworks. As illustrated in subfigure (a), QSC-Net gives the best Secure Packet Delivery of 96.8%, followed by SDN-IDS (82.4%),  BB84-QKD (87.9%) and Secure IoT Gateway (85.1%). Under hostile network conditions where packet loss and replay attacks may occur, the system uses quantum-based encryption, combined with federated trust verification, to successfully defeat both threats. In Subfigure (b), Trust Establishment Time is illustrated, where QSC-Net accomplishes a minimum of 22.3 ms latency thanks to its compact Q-PUF and the credential exchange protocol adaptable to the new peer. Secure IoT Gateway takes 49.2ms to verify identity using a conventional RSA-based method; the time consumed by BB84-QKD and SDN-IDS for classical handshaking is 35.6 ms and 40.4 ms, respectively. Subfigure (c) illustrates Attack Response Latency, where QSC-Net provides low-latency detection and mitigation with a 11.2 ms response time. As a result, SDN-IDS, BB84-QKD, and Secure IoT Gateway show response times of 23.7 ms, 19.1 ms, and 25.6 ms, respectively; the system’s performance largely owes to the federated anomaly detection system, fed via policy update methods that learn directly via reinforcement learning. Subfigure (d) depicts the Throughput Under Attack metric, showing that QSC-Netat achieves 78.6 Mbps, exceeding BB84-QKD (64.5 Mbps), Secure IoT Gateway (58.3 Mbps), and SDN-IDS (61.7 Mbps). It self-protects against bandwidth attacks by incorporating adaptive encryption and quantum-aware routing. The graph in subfigure(e) shows System Scalability, the ability for the number of nodes to scale without hurting performance. The QSC-Net system is scalable and uses 500 nodes (which exceed the maximum node counts of 350 for the BB84-QKD system, 280 for the Secure IoT Gateway, and 300 for SDN-IDS). The upcoming system will offer greater scalability through its modular, distributed architecture, which comprises quantum-assisted coordination systems. From Fig. 8 , we observe that QSC-Net outperforms Larbi et al.‘s method across transmission security, latency, response time, resilience, and scalability, indicating that it is the best secure communication system for connected smart cities. Discussion on resilience, scalability, and overhead As shown in previous sections, the QSC-Net framework has demonstrated its ability to protect communication in smart cities. This system provides remarkable resilience against a wide variety of threats, including DDoS attacks, spoofing, and quantum-specific threats such as photon siphoning and QKD link eavesdropping. By combining quantum encryption via QKD with lightweight PUF-based identity authentication and federated anomaly detection (which can ensure an F1-score above 93% for threat classification under privacy-preserving conditions), the system provides resilience against security threats. In the scalability analysis of QSC-Net, we observed that its performance overhead increases linearly as the system scales up to additional nodes. The core impediment to both the Secure IoT Gateway and SDN-IDS systems is reliance on centralised processing, though performance starts to suffer around 300–350 nodes. In contrast, QSC-Net runs smoothly even with 500 nodes, thanks to federated learning and distributed trust validation. Experimental results show that the RL-based routing engine outperforms the others, adapts well to the dynamic nature of a mobile ad hoc network topology, and converges rapidly with a high packet delivery ratio. When we compare quantum and classical layers, we can see multiple trade-offs that affect the performance of these two types of layers. Although these QKD modules encrypt keys in a practically unbreakable way, they also require additional time for the key exchange and contribute to photon loss, which has been shown to cause high QBER (quantum bit error rate) in simulations. Experimental studies report that, at most 60 km of transmission distance, no error correction is required for QKD links, but more energy is consumed. This comparison shows that classical encryption methods, such as Kyber, are faster but will be vulnerable to quantum attacks. The hybrid QSC-Net effortlessly balances these trade-offs by operating with either quantum or post-quantum encryption, switching between them based on link quality and network needs. Increased the data transmission requirements a bit due to the synchronisation process between the FL models and the node-to-node synchronisation, and DP noise. That might seem like a high processing cost, but they save critical infrastructure systems—and health and personal mobility information. Updates for FL with DP incur ∼7% more communication overhead while providing full end-to-end privacy against any middleman attacks. The practical deployment yields promising outcomes, which are conducive to a positive assessment of its feasibility. QSC-Net uses a modular architecture that enables the user to plug in separate QKD, Q-PUF, RL-routing, and FL-detection systems. The SDN controller and MEC server are integrated to enforce edge-level security policies more directly. The three most significant challenges QKD systems face are hardware scalability, suppression of environmental quantum noise, and the preservation of Q-PUF entropy over large networks of devices. QSC-Net introduces a modern paradigm for resilient and secure urban communication systems that leverages quantum features, cutting overhead and balancing the resilience, scalability, and security trade-off. Time complexity and system overhead analysis Due to the stringent real-time requirements of smart city communication systems, it is important to quantify the computational and communication overhead imposed by the proposed QSC-Net components. This subsection provides an analysis of theoretical time complexity and an overhead evaluation based on empirical measurements of the core modules: RL-based secure routing, federated anomaly detection, quantum key distribution (QKD), and Q-PUF-based authentication. RL-based secure routing complexity Let denote the state vector dimension, the number of routing actions, the number of neighboring nodes for node , and the number of hidden neurons in the deep Q-network (DQN). For each packet forwarding decision, the RL agent performs a forward pass through the DQN with complexity . Neighbor reliability computation and next-hop selection incur an additional cost of . Hence, the total per-packet routing decision complexity is Since model training is performed offline or at coarse time intervals, the online routing path remains lightweight and suitable for real-time operation. Federated learning aggregation complexity Assume participating nodes, model parameters, local epochs, and samples at node . Local model training at each node has complexity , while server-side aggregation using FedAvg incurs a complexity of per round. Communication overhead per round scales linearly as , since only model parameters or gradients are exchanged. This linear scaling ensures that the federated anomaly detection module remains practical for large-scale smart city deployments. Quantum key distribution and Q-PUF authentication overhead The time required for QKD depends on session initialization, key sifting, error reconciliation, and privacy amplification. The overall QKD session cost can be approximated as where is the generated key length and is the secure key rate. When quantum channel quality degrades, QSC-Net seamlessly switches to post-quantum cryptography, whose encryption overhead scales linearly with data size. For Q-PUF authentication, challenge–response generation and verification require constant time, where cap Ls is he response length, resulting in negligible latency compared to cryptographic operations. Empirical overhead and scalability evaluation The theoretical analysis was validated using empirical measurements across node-count variations in small-scale to large-scale imaginative city scenarios. They demonstrate that in RL routing inference, latency is still in the millisecond range, that federated aggregation time scales linearly with node count, and that QKD/Q-PUF operations incur slight per-node overhead relative to the total communication delay. The results confirm that QSC-Net meets the requirements of real-time communication while scaling well with network size and workload intensity. The runtime overhead of essential components in QSC-Net with varying numbers of participating nodes is shown in Table 10 . RL-based Routing Inference Latency, federated learning aggregation time, QKD session setup overhead, and Q-PUF authentication delay are reported. The results show quasi-linear scalability with respect to network size, while security-related operations incur bounded, practically reasonable latency for real-time smart city communications. Table 10. Runtime overhead scaling with number of nodes. Number of nodes ( N ) RL routing inference time (ms) Federated aggregation time per round (s) QKD Session setup time (ms) Q-PUF authentication time (ms) 50 0.9 ± 0.1 0.42 ± 0.05 18.3 ± 2.1 1.6 ± 0.2 100 1.2 ± 0.2 0.78 ± 0.07 19.6 ± 2.4 1.7 ± 0.2 200 1.6 ± 0.2 1.41 ± 0.11 21.2 ± 2.8 1.9 ± 0.3 500 2.3 ± 0.3 3.12 ± 0.26 23.9 ± 3.1 2.1 ± 0.3 Open in a new tab A unified runtime scalability analysis of key QSC-Net components as the number of participating intelligent city nodes increases is presented in Fig. 10 . Results: The combined plot presents the inference latency of the RL-based secure routing module, the per-round aggregation time of the federated anomaly detection phase, the setup overhead for quantum key distribution (QKD) sessions, and the authentication-induced delay caused by the Q-PUF mechanisms. We observe near-linear scaling with network size, confirming the practical viability of deploying QSC-Net in large-scale urban IoT environments. Most critically, security-critical operations incur bounded, predictable overhead and maintain real-time guarantees while providing strong, quantum-resilient security across heterogeneous smart city domains. Fig. 10. Open in a new tab Runtime scalability analysis of QSC-Net components under increasing smart city network size. Comparison with AI-based and quantum-resilient security frameworks While these baseline comparisons with SDN-IDS, BB84-QKD, and secure IoT gateway approaches confirm the validity of QSC-Net under classical and quantum-only requirements, the focus of smart-city security research shifts towards learning-driven adaptation and quantum-resilient cryptographic designs. To improve external validity, we expand the evaluation to include representative strong state-of-the-art categories: (a) PQC-only secure communication (which provides for post-quantum key exchange and symmetric encryption, but without QKD), (b) Hybrid QKD–PQC without AI (link-level quantum/classical switching but without adaptive routing or distributed detection), (c) FL-IDS baseline (federated anomaly detection without quantum-secure keying and without routing adaptation), and (d) RL-secure routing baseline (learning-based routing with classical keying and without federated security analytics). To make a fair comparison, these baselines are implemented in software/simulation using the same network and traffic profiles as those in the PHE-EO. Our results demonstrate the unmatched advantage of QSC-Net, which end-to-end integrates quantum-resilient keying, trust-aware RL routing, and privacy-preserving federated detection to jointly enhance delivery security, resilience against attacks, and scalability across both smart-city mobility and cross-domain traffic. *Runtime overhead denotes the additional per-decision (routing/auth/security) processing delay measured at the node level (mean ± std). As shown in Table 11 , single-stream strategies (PQC-only, hybrid QKD–PQC w/o AI, FL-only, or RL-only) improve a subset of objectives but do not yield consistent end-to-end gains for smart-city dynamics. PQC-only methods reduce quantum dependence, yet are link-agnostic for secrecy and lack threat-adaptive integration, resulting in limited secure delivery. Hybrid QKD–PQC without AI increases key robustness but lacks responses to congestion or adversarial routing behaviour. Trust degradation influences routing, and the FL-IDS method improves detection but not route optimisation. Although RL secure routing improves both latency and delivery, it does not provide anomaly intelligence in a distributed, privacy-preserving manner. On the contrary, QSC-Net yields the best secure delivery, latency, and detection F1 in the long term, since it tightly integrates quantum-resilient keying with real-time RL routing and local federated anomaly detection, enabling coordinated adaptation at scale across communication, trust, and security layers. Table 11. Comparison with recent AI-based and quantum-resilient baselines ( N = 500 nodes). Method Security stack Adaptation mechanism Secure packet delivery ratio (%) Avg end-to-end latency (ms) Intrusion detection F1-score Runtime overhead (ms)* QSC-Net (proposed) Hybrid QKD–PQC + Q-PUF + FL RL routing + FL aggregation 96.8 ± 1.1 42.6 ± 4.3 0.94 ± 0.02 2.3 ± 0.3 PQC-only secure IoT PQC key exchange + symmetric crypto Static policies 90.7 ± 1.8 49.8 ± 5.1 0.86 ± 0.03 1.7 ± 0.2 Hybrid QKD–PQC (No AI) QKD–PQC switching Link-only switching 92.4 ± 1.6 51.2 ± 5.4 0.83 ± 0.04 2.0 ± 0.3 FL-IDS Baseline Classical crypto + FL IDS Federated only 93.1 ± 1.5 47.9 ± 4.9 0.91 ± 0.02 2.1 ± 0.3 RL secure routing baseline Classical crypto RL routing only 94.0 ± 1.4 45.7 ± 4.6 0.85 ± 0.03 2.2 ± 0.3 SDN-IDS baseline SDN monitoring + IDS Centralized control 91.6 ± 1.7 53.4 ± 5.7 0.88 ± 0.03 2.6 ± 0.4 BB84-QKD baseline QKD keying only No learning 89.3 ± 2.0 58.1 ± 6.0 0.81 ± 0.04 2.4 ± 0.4 Open in a new tab As shown in Fig. 11 , a comparative analysis of QSC-Net is performed against representative state-of-the-art AI-aware and quantum-resilient security frameworks, such as PQC-only secure IoT communication, hybrid QKD–PQC without adaptive intelligence, federated learning–based intrusion detection, reinforcement learning–based secure routing, SDN-IDS, and BB84-QKD approaches. Comparison of smart-city networks under the same conditions on secure packet delivery ratio, end-to-end latency, and intrusion detection accuracy. Our results show that, by combining hybrid quantum-classical keying, adaptive RL-driven routing, and privacy-preserving federated anomaly detection, QSC-Net outperforms all baselines in terms of security, scalability, and real-time performance in large-scale urban IoT scenarios. Fig. 11. Open in a new tab Comparative performance evaluation of QSC-Net against recent AI-based and quantum-resilient security frameworks. Statistical significance analysis Multi-independent runs were compared using non-parametric statistical tests (to check if reported improvements in performance were statistically significant). All of the rolling horizon algorithms have been implemented for 10 independent trials with different random seeds, and the results are presented as mean ± standard deviation. We first applied a Friedman test to determine whether there were statistically significant differences among the competing algorithms. Then, a pairwise Wilcoxon signed-rank test was conducted between the proposed MF-MBO framework and each of the baseline algorithms at the 95% confidence level, with a p-value of ). In addition, 95% confidence intervals were computed for key performance metrics to assess the stability and robustness of the results. These analyses confirm that the observed improvements of MF-MBO are not due to random variation and are statistically reliable. We performed all tests using the non-parametric methods appropriate for stochastic optimisation algorithms. Table 12 Statistical Significance on Comparison of MF-MBO Framework against Baseline Algorithms using Non-Parametric Tests. The Friedman and Wilcoxon signed-rank test results we now report (and the significance decisions at the 95% confidence level) show that the performance improvements observed with MF-MBO (over DM-MBO) are statistically significant and not due to random variation across multiple experimental runs. Table 12. Statistical significance analysis of MF-MBO compared with baseline algorithms. Comparison Friedman test ( p -value) Wilcoxon test ( p -value) Statistically Significant (α = 0.05) MF-MBO vs. GA < 0.001 0.003 Yes MF-MBO vs. PSO < 0.001 0.005 Yes MF-MBO vs. MBO < 0.001 0.008 Yes Open in a new tab Convergence analysis of MF-MBO and baseline algorithms To enhance the transparency and readability, this subsection displays the convergence procedure of the described MF-MBO alongside competing baseline algorithms. Beyond final performance values, the convergence curves provide insight into the optimisation dynamics and stability of each method. The same stopping criteria and population size were used for all algorithms, and the best objective value was recorded at each run iteration Fig. 12 . Fig. 12. Open in a new tab Convergence analysis of MF-MBO and baseline optimisation algorithms. The convergence trends for MF-MBO, GA, PSO, and standard MBO with respect to the main optimisation objective are presented in Fig. 12 . These results indicate that more exploration is achieved in early iterations with MF-MBO through global fuzzy dominance–based selection and the quantum-inspired tunnelling mechanism. The greedy migration strategy enables finer exploitation in later generations, leading to a faster stabilising process and better solutions. Model 1: In comparison, baseline algorithms take longer to converge and are also more susceptible to local optima. These observations confirm that MF-MBO not only converges faster but also remains stable across multiple runs. Ablation study of the MF-MBO framework An ablation study was conducted to assess the contribution of each design component in the proposed MF-MBO framework across all identical experimental settings described in the main evaluation. This study aims to isolate and measure the influence of fuzzy dominance modelling, quantum-inspired tunnelling, and greedy migration refinement on system performance. The exact population size, iteration budget, workload configuration, and random seed protocol discussed in Sect.  5.1 and 5.4 were used for all the ablated variants. Results are expressed as mean ± standard deviation across 10 independent simulations for each configuration. Four different MF-MBO variants were evaluated: (i) Full MF-MBO, which employs fuzzy dominance, tunnelling, and greedy migration; (ii) MF-MBO without Fuzzy Dominance, wherein the ranking of the solutions depends on the traditional weighted fitness aggregation; (iii) MF-MBO without Tunnelling, wherein non-dominant solutions are never accepted with a certain probability, thus leading to purely greedy evolution; and (iv) MF-MBO without Greedy Migration, wherein local refinement and load-balancing adjustments are disabled. As shown in Table 13 , removing any component significantly impairs performance across all metrics. Specifically, deactivating fuzzy dominance increases multi-objective decision instability, and eliminating tunnelling leads to changes in premature convergence and higher latency. Without greedy migration, load balancing deteriorates, leading to higher energy consumption and greater runtime overhead. These results validate that the performance improvements achieved by MF-MBO result from the combined effect of all three mechanisms, not from any single heuristic. Table 13. Ablation study of MF-MBO components (Mean ± Std over 10 Runs). MF-MBO variant End-to-End latency (ms) ↓ Energy consumption (J) ↓ Packet Delivery ratio (%) ↑ Runtime overhead (ms) ↓ Full MF-MBO 41.6 ± 1.9 1.72 ± 0.05 96.8 ± 0.7 18.4 ± 0.8 w/o fuzzy dominance 46.9 ± 2.3 1.89 ± 0.06 93.4 ± 0.9 21.6 ± 1.0 w/o tunneling 49.7 ± 2.6 1.94 ± 0.07 92.1 ± 1.1 20.9 ± 0.9 w/o greedy migration 47.8 ± 2.4 1.88 ± 0.06 93.0 ± 1.0 22.3 ± 1.1 Open in a new tab Comparison with existing methods A detailed comparison of QSC-Net with several quantum and classical security schemes not only provides a basis for evaluating QSC-Net’s effectiveness but also reveals differences in design range, characteristics, and security performance. Due to the focus on specific aspects, a comprehensive integration of authentication and encryption, anomaly detection, and federated learning has not yet been addressed within a single, quantum-secure, adaptive framework in the existing work. Javeed et al. They have created a federated system based on quantum cryptography that serves as a 6G-IoT application platform 1 . It locks down model updates (based on the quantum method), but provides no corresponding path at present and no explanation of the components. QSC-Net: A Federated Learning based Explainable Anomaly Detection System using Reinforcement Learning. Shabbir et al. A fuzzy-logic-based quantum authentication system for smart city 19 . While it has more than one step to confirm a user’s identity, it doesn´t work well because users are changing places. The QSC-Net system addresses these challenges with Q-PUF at its core and federated trust layers that address scalability challenges hindering the global adoption of quantum-secure connected systems. Barletta et al. introduced a hybrid encryption system that utilises both QKD and classical cryptographic systems in 17 . For example, the system does not operate with any built-in routing intelligence or machine-mediated anomaly detection. In contrast, we address this by developing a novel RL-based, federated, and privacy-preserving detection system in QSC-Net. We qualitatively compare QSC-Net with existing methods; the comparative results are shown in Tab. Khayyat 68 developed an edge-based intrusion detection system by combining deep learning methods with bio-inspired optimisation techniques. The solution provides neither privacy protection nor quantum security because QSC-Net uses a differentially private federated module that employs QKD-secured updates. Kalinin and Krundyshev 61 used quantum-enhanced machine learning for IDS. Their system achieves strong detection performance but lacks interpretability and fails to protect user privacy. QSC-Net provides interpretable outputs with secure aggregation. Senor et al. 53 and Karbasi and Shahpasand 48 developed lightweight post-quantum cryptographic systems through their work on NTRU and simplified AES implementations. These systems operate efficiently, but their functionality remains restricted to device-level encryption because they remain vulnerable to adaptive attack methods. QSC-Net operates through layered QKD encryption alongside Q-PUF authentication and routing defence mechanisms. Table 10 presents a qualitative feature-based comparison of QSC-Net with other techniques across eight fundamental quantum-security capabilities. Table 14 . Table 14. Qualitative comparison of QSC-Net with existing methods. Method / reference QKD Support RL-based routing FL + Privacy Anomaly detection PUF-based Auth Explainability Real-time operation Smart city scope QSC-Net (proposed) ✓ ✓ ✓ (DP-FL) ✓ (Explainable) ✓ ✓ ✓ ✓ Javeed et al. 1 ✓ ✗ ✓ ✗ ✗ ✗ ✗ ✓ Shabbir et al. 19 ✓ (Fuzzy) ✗ ✗ ✗ ✓ ✗ ✗ ✓ Barletta et al. 17 ✓ ✗ ✗ ✗ ✗ ✗ ✗ ✓ Khayyat 68 ✗ ✗ ✗ ✓ (DL) ✗ ✗ ✓ ✗ Kalinin & Krundyshev 61 ✓ (QML) ✗ ✗ ✓ ✗ ✗ ✗ ✗ Senor et al. 53 ✗ ✗ ✗ ✗ ✗ ✗ ✓ ✓ Karbasi & Shahpasand 48 ✗ ✗ ✗ ✗ ✗ ✗ ✓ ✗ Cao et al. 54 ✓ ✗ ✗ ✗ ✗ ✗ ✓ ✗ Rishiwal et al. 56 ✗ ✗ ✗ ✗ ✓ ✗ ✓ ✓ Open in a new tab Cao et al. 54 presented QKD-based grid systems. Despite the strong smart-grid communication features the solution offers, it applies only to specific domains and therefore cannot support multiple heterogeneous, innovative city environments 19 . QSC_NET: It’s planned to connect multiple UPS sets. While previous studies focused on individual components, our QSC-Net offers a unique combination of quantum-resistant encryption, adaptive RL-routing, privacy-aware anomaly detection, and PUF-based federated identity, and we achieve its excellent performance by integrating these components. The performance comparison depicted in Fig. 13 shows that QSC-Net provides a specific set of abilities needed to implement a scalable, real-time, quantum-resistant smart city security system that outperforms existing solutions since they lack both flexibility and decoupled learning, and require a dedicated approach to secure real-time data exchange between the cloud and user and entity-facing devices. Fig. 13. Open in a new tab Qualitative comparison of QSC-Net and existing methods. Discussion Innovative city ecosystems are becoming complex systems in which decision-making systems need to operate with intelligent behaviours, high efficiency, and adapt to changing environmental conditions. While conventional AI systems provide low-level automation and optimisation, they cannot sustain it in dynamic city-scale environments. This new research integrates quantum-inspired optimisation principles with deep learning methodologies to develop a framework that will support sustainable AI services in a smart city. This state-of-the-art survey shows that existing models rely on static neural network designs and hard-coded algorithms, offering limited adaptability and failing to account for power consumption. Current models treat sustainability more as a minor afterthought than as an essential design component. This work reveals a significant gap between existing studies. It proposes a quantum-inspired algorithm that speeds up calculations in an environmentally friendly way, with sustainability as its primary optimisation target. By integrating both Amplitude-Encoded Quantum Optimisation and Energy-Aware Neural Architecture Search (E-NAS) and Green Scheduling systems, we establish a new methodology. Together, the components lower training power consumption, better manage inference resources, and lay the foundation for reduced model instability in UgoChip. Our quantum-inspired system demonstrates significantly higher performance than classical baselines, with reduced embedded carbon emissions, faster convergence, and higher task effectiveness under limited power constraints. Our evaluation shows that Q-Enhance surpasses conventional NAS and pruning approaches in terms of accuracy, energy efficiency, and model complexity control. It addresses several current challenges related to existing analytical methods. Overparameterization, static assignment of hardware resources, and static retraining cycles combine to make traditional deep learning models inefficient and non-adaptive. Yet the new framework builds an adaptive learning and deployment process to produce low-structure but high-performance models, simply using quantum-inspired tunnelling and reinforcement-driven controller optimisation. Although scheduling mechanisms that account for carbon intensity levels exist to avoid compromising the system’s sustainability, they have not yet been developed. The research results provide multiple insights into the mentioned fields. The research provides a framework that combines deep learning, quantum-inspired and eco-design to enable AI to be deployed for vital operations in smart cities — like energy management, dynamic traffic management, and sustainable urban planning — without hindering environmental goals. This system sets a new standard for developing scalable and accurate AI models, as well as for operationalising them efficiently, responsibly and sustainably. The system boundaries are documented in Sect.  6.1, as are methods for expanding its capacity to manage diverse domains and for future hardware upgrades. Limitations of the study While this quantum-assisted system is more energy-efficient, faster, and more versatile than today’s systems, its existing limitations will require further study by the researchers. The current implementation and evaluation are performed using a simulated task orchestration environment for the smart city. Simulators work as expected; however, operational environments introduce additional constraints, such as other hardware devices, variable network speeds, and signal interference, which the registrar does not account for. Second, the tunnelling and amplitude reweighting modules, which are the core of our quantum-inspired components, are designed using approximations rather than a direct mapping to quantum hardware. Specifically, the research team created a method to model quantum operations on classical computers. However, they have not yet tested a physical implementation on an NISQ device, a step that could reveal additional challenges in stability, latency, and scalability in their model. Third, while the framework shows over 90% resilience and scalability across tasks and agents under typical workloads, in highly time-sensitive scenarios such as disaster response or traffic control, the dynamic reconfiguration overhead may become non-negligible, requiring benchmarking of this overhead as a first-class task. The system’s ability to adapt to variability does not keep pace with the time needed to reconfigure during extreme events. The limitations set boundaries for future studies. In fact, Sect.  6.1 of the paper contains a detailed discussion of the system’s vulnerabilities and possible remedies. Conclusion and future work This paper proposes a quantum-based fuzzy optimisation system as a novel city cloud task-scheduling system. In this work, we present a new method that uses fuzzy dominance-based decision-making with a hybrid Monarch Butterfly Optimisation system to achieve dynamic load balancing, energy minimisation, and execution timing optimisation. This detailed task-resource matching enables faster task execution and lower power consumption, resulting in a reduced make span in benchmarks compared to typical approaches, as shown by results using the new method. By enhancing dominance during migration and local search operations with fuzzy reaches, the proposed method enables better candidate selection under high system load. The implementation was validated experimentally using a set of tasks and different types of VMs, and the performance model was shown to reliably predict system observables, resulting in a 12% improvement in energy efficiency and a 14% reduction in make span compared to baseline models. Results from the evaluation show that the proposed method is well-suited to real-world imaginative city scenarios, as demonstrated by actual testbeds that require fast-acting systems, fair resource allocation, and high-performance computation. It works fine with fixed infrastructure, but with the same VMs, it does not handle node failures or account for changing operational costs. Section  6.1 discusses such constraints. The evolution of such a framework will be able to handle heterogeneous environments, support real-time task processing, and be multi-objective, tackling multi-criteria decision-making problems regarding cost, environmentally friendly operations, and scheduling, making it capable of trust-based operations. Adaptive learning from the system, in conjunction with edge-cloud coordination systems, will enable the system to acquire enhanced functionality for smart city operations. Author contributions All authors contributed to the study’s conception and design. Material preparation, data collection, and analysis were performed by Dr. Nalavala Ramanjaneya Reddy, Dr. G Arul Dalton, K Swathi, Dasaka VSS Subrahmanyam, Lakshmi Kranthi G, Gujjeti Nagaraju. The first draft of the manuscript was written by Dr. Nalavala Ramanjaneya Reddy all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Data availability Data is available with the corresponding author and will be given on request. Materials availability Materials used in this research are available with corresponding author and given on request. Code availability The code is available from the corresponding author and can be given on request. Declarations Competing interests The authors declare no competing interests. Consent for publication The authors give consent for their publication. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. Javeed, D. et al. Quantum-empowered federated learning and 6G wireless networks for IoT security: Concept, challenges and future directions. Future Gener Comput. Syst. 160 , 577–597 (2024). [ Google Scholar ] 2. Muheidat, F., Dajani, K. & Tawalbeh, L. A. Security concerns for 5G/6G mobile network technology and quantum communication. Procedia Comput. Sci. 203 , 32–40 (2022). [ Google Scholar ] 3. Hsu, C. H. et al. Green communication approach for the smart city using renewable energy systems. Energy Rep. 8 , 9528–9540 (2022). [ Google Scholar ] 4. Chawla, D. & Mehra, P. S. A survey on quantum computing for Internet of Things security. Procedia Comput. Sci. 218 , 2191–2200 (2023). [ Google Scholar ] 5. Akbar, M. A., Khan, A. A. & Hyrynsalmi, S. Role of quantum computing in shaping the future of 6G technology. Inf. Softw. Technol. 170 , 107452 (2024). [ Google Scholar ] 6. Iftikhar, A., Qureshi, K. N., Shiraz, M. & Albahli, S. Security, trust and privacy risks, responses, and solutions for high-speed smart cities networks: A systematic literature review. J. King Saud Univ. - Comput. Inf. Sci. 35 (10), 101794 (2023). [ Google Scholar ] 7. Ahad, M. A., Paiva, S., Tripathi, G. & Feroz, N. Enabling technologies and sustainable smart cities. Sustain. Cities Soc. 61 , 102301 (2020). [ Google Scholar ] 8. Kumar, A., Pacheco, D. A., de Kaushik, J., Rodrigues, J. J. P. C. & K. and Futuristic view of the Internet of Quantum Drones: Review, challenges and research agenda. Veh. Commun. 36 , 100487 (2022). [ Google Scholar ] 9. Chen, Z., Gan, W., Wu, J., Lin, H. & Chen, C. M. Metaverse for smart cities: A survey. Internet Things Cyber-Phys Syst. 4 , 203–216 (2024). [ Google Scholar ] 10. Qadir, Z., Ullah, F., Munawar, H. S. & Al-Turjman, F. Addressing disasters in smart cities through UAVs path planning and 5G communications: A systematic review. Comput. Commun. 168 , 114–135 (2021). [ Google Scholar ] 11. Singh, P. R., Singh, V. K., Yadav, R. & Chaurasia, S. N. 6G networks for artificial intelligence-enabled smart cities applications: A scoping review. Telemat Inf. Rep. 9 , 100044 (2023). [ Google Scholar ] 12. Akhtar, M. W. et al. The shift to 6G communications: Vision and requirements. Hum. -Cent Comput. Inf. Sci. 10 , 1–27 (2020). [ Google Scholar ] 13. Hosseinzadeh, M., Hemmati, A. & Rahmani, A. M. Clustering for smart cities in the Internet of Things: A review. Clust Comput. 25 (6), 4097–4127 (2022). [ Google Scholar ] 14. Ali, J. & Zafar, M. H. Improved end-to-end service assurance and mathematical modeling of message queuing telemetry transport protocol based applications in smart cities. Alex Eng. J. 72 , 657–672 (2023). [ Google Scholar ] 15. Wang, M. et al. Security and privacy in 6G networks: New areas and new challenges. Digit. Commun. Netw. 6 (3), 281–291 (2020). [ Google Scholar ] 16. Qadir, Z., Le, K. N., Saeed, N. & Munawar, H. S. Towards 6G Internet of Things: Recent advances, use cases, and open challenges. ICT Express . 9 (2), 296–312 (2023). [ Google Scholar ] 17. Barletta, V. S., Caivano, D., De Vincentiis, M., Pal, A. & Scalera, M. Hybrid quantum architecture for smart city security. J. Syst. Softw. 217 , 112161 (2024). [ Google Scholar ] 18. Zhou, Y. et al. Quantum computing in power systems. iEnergy 1 (2), 170–187. (2022). 19. Shabbir, M., Ahmad, F., Shabbir, A. & Alanazi, S. A. Cognitively managed multi-level authentication for security using fuzzy-logic-based quantum key distribution. J. King Saud Univ. - Comput. Inf. Sci. 34 (4), 1468–1485 (2022). [ Google Scholar ] 20. Chiti, F., Picchi, R. & Pierucci, L. A survey on non-terrestrial quantum networking: Challenges and trends. Comput. Netw. 252 , 110619 (2024). [ Google Scholar ] 21. Manimuthu, A., Dharshini, V., Zografopoulos, I., Suryanarayanan, S. & Venkatesan, R. Contactless technologies for smart cities: Big data, IoT, and cloud infrastructures. SN Comput. Sci. 2 , 1–24 (2021). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Tightiz, L., Dang, L. M., Padmanaban, S. & Hur, K. Metaverse-driven smart grid architecture. Energy Rep. 12 , 2014–2025 (2024). [ Google Scholar ] 23. Gill, S. S., Wu, H., Patros, P., Ottaviani, C. & Prieto-González, L. Modern computing: Vision and challenges. Telemat Inf. Rep. 13 , 100116 (2024). [ Google Scholar ] 24. Gad, A. G., Mosa, D. T., Abualigah, L. & Abohany, A. A. Emerging trends in blockchain technology and applications: A review and outlook. J. King Saud Univ. - Comput. Inf. Sci. 34 (9), 6719–6742 (2022). [ Google Scholar ] 25. Gourisetti, S. N. G., Cali, Ü., Choo, K. K. R. & Elias, D. Standardization of the distributed ledger technology cybersecurity stack for power and energy applications. Sustain. Energy Grids Netw. 28 , 100553 (2021). [ Google Scholar ] 26. Shi, W., Goodchild, M., Batty, M., Li, Q. & Liu, X. Prospective for urban informatics. Urban Inf. 1 , 1–14 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Yi, H. Improving cloud storage and privacy security for digital-twin-based medical records. J. Cloud Comput. 12 , 1–16 (2023). [ Google Scholar ] 28. Marengo, A. Navigating the nexus of AI and IoT: A comprehensive review of data analytics and privacy paradigms. Internet Things . 27 , 101279 (2024). [ Google Scholar ] 29. Vanderhorst, H. R., Heesom, D. & Yenneti, K. Technological advancements and the vision of a meta smart twin city. Technol. Soc. 79 , 102731 (2024). [ Google Scholar ] 30. Seedorf, J. et al. A prototype for evaluating post-quantum cryptography on resource-constrained hardware with real-world smart city sensor data. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLVIII-4/W16-2025, 113–120. 10.5194/isprs-archives-XLVIII-4-W16-2025-113-2025 (2025). 31. Javed, M. A. et al. Quantum computing-assisted 6G networks: Use cases and future opportunities. Tsinghua Sci. Technol. 10.26599/TST.2024.9010118 (2025). [ Google Scholar ] 32. Yalcin, H., Daim, T., Mokhtari Moughari, M. & Mermoud, A. Supercomputers and quantum computing on the axis of cybersecurity. Technol. Soc. 77 , 102523 (2024). [ Google Scholar ] 33. Trincado-Munoz, F., van Meeteren, M., Rubin, T. H. & Timmerman, J. Digital transformation in the world city networks’ advanced producer services complex: A technology space analysis. Geoforum 151 , 104001 (2024). [ Google Scholar ] 34. Admass, W. S., Munaye, Y. Y. & Diro, A. A. Cybersecurity: State of the art, challenges and future directions. Cyber Secur. Appl. 2 , 100046 (2024). [ Google Scholar ] 35. Bokolo, A. J., Sylva, W., Kuika Watat, J. & Misra, S. A framework for standardization of distributed ledger technologies for interoperable data integration and alignment. J. Knowl. Econ. 15 , 12053–12096 (2024). [ Google Scholar ] 36. Vaigandla, K. K. Quantum-secure IoT networks for the 6G era: Post-quantum cryptography, blockchain integration, and trust architectures - A comprehensive review. J. Sens. IoT Health Sci. 3 (3). 10.69996/jsihs.2025014 (2025). 37. Fischer-Hübner, S. & Alcaraz, C. Stakeholder perspectives and requirements on cybersecurity in Europe. J. Inf. Secur. Appl. 61 , 102916 (2021). [ Google Scholar ] 38. Alharbi, S. QuantumTrust-FedChain: A blockchain-aware quantum-tuned federated learning system for cyber-resilient industrial IoT in 6G. Future Internet . 17 (11), 493. 10.3390/fi17110493 (2025). [ Google Scholar ] 39. Chen, S. J. & Tsai, Y. H. Quantum-safe networks for 6G: An integrated survey on PQC, QKD, and satellite QKD with future perspectives. Comput. AI Connect. 2 , 0016 10.69709/CAIC.2025.102135 (2025). 40. Schmitt, M. Securing the digital world: Protecting smart infrastructures and digital industries with artificial intelligence (AI). J. Ind. Inf. Integr. 36 , 100511 (2023). [ Google Scholar ] 41. Gupta, S., Roy, D., Pramanik, P. & Chowdhury, S. Enhancing cyber security in 6G networks with federated learning for collaborative threat mitigation. In: 6G Cyber Security Resilience: Trends and Challenges . (Springer, 2025). 42. Manulis, M., Bridges, C. P., Harrison, R., Sekar, V. & Davis, A. Analysis of threats, key enabling technologies and challenges. Int. J. Inf. Secur. 20 , 287–311 (2021). [ Google Scholar ] 43. Rejeb, A., Rejeb, K., Appolloni, A., Jagtap, S. & Mangla, S. K. Unleashing the power of Internet of Things and blockchain: A comprehensive analysis and future directions. Internet Things Cyber-Phys Syst. 4 , 1–18 (2024). [ Google Scholar ] 44. Ahad, A., Ali, Z., Mateen, A., Tahir, M. & Hannan, A. A comprehensive review on 5G-based smart healthcare network security: Taxonomy, issues, solutions and future research directions. Array 18 , 100279 (2023). [ Google Scholar ] 45. Mahdi, L. H. & Abdullah, A. A. Fortifying future IoT security: A comprehensive review on lightweight post-quantum cryptography. Eng. Technol. Appl. Sci. Res. 15 (2), 21812–21821 (2025). [ Google Scholar ] 46. Abbas, A. M., Youssef, K. Y., Mahmoud, I. I. & Zaki, F. W. NB-IoT optimization for smart meters networks of smart cities: Case study. Alex Eng. J. 59 (6), 4267–4281 (2020). [ Google Scholar ] 47. Durr-e-Shahwar, Imran, M., AlTamimi, A. B., Khan, W. & Sharif, A. Quantum cryptography for future networks security: A systematic review. IEEE Access. 12 , 1–31 (2024). [ Google Scholar ] 48. Karbasi, A. H. & Shahpasand, S. SINGLETON: A lightweight and secure end-to-end encryption protocol for the sensor networks in the Internet of Things. J. Supercomput . 76 , 1–39 (2020). [ Google Scholar ] 49. Kumari, S., Singh, M., Singh, R. & Tewari, H. To secure the communication in powerful Internet of Things using innovative post-quantum cryptographic method. Arab. J. Sci. Eng. 46 , 1–17 (2021).33173717 [ Google Scholar ] 50. Pillai, S. E. V. S. & Polimetla, K. Analyzing the impact of quantum cryptography on network security. IEEE Access. 12 , 1–7 (2024). [ Google Scholar ] 51. Singh, A., Dev, K., Siljak, H., Hossain, M. S. & Ravi, V. Quantum Internet—Applications, functionalities, enabling technologies, challenges, and research directions. IEEE Commun. Surv. Tutor. 23 (4), 2218–2247 (2021). [ Google Scholar ] 52. Muniswamy, A. & Rathi, R. A detailed review on enhancing the security in Internet of Things-based smart city environment using machine learning algorithms. IEEE Access. 12 , 120389–120413 (2024). [ Google Scholar ] 53. Señor, J., Portilla, J. & Mujica, G. Analysis of the NTRU post-quantum cryptographic scheme in constrained IoT edge devices. IEEE Internet Things J. 9 (20), 18778–18970 (2022). [ Google Scholar ] 54. Cao, Y. et al. The evolution of quantum key distribution networks: On the road to the Qinternet. IEEE Commun. Surv. Tutor. 24 (4), 1–59 (2022). [ Google Scholar ] 55. Rathee, G., Iqbal, R., Waqar, O. & Guizani, M. On the design and implementation of a blockchain-enabled e-voting application within IoT-oriented smart cities. IEEE Access. 9 , 34165–34176 (2021). [ Google Scholar ] 56. Rishiwal, V., Agarwal, U. & Alotaibi, A. Exploring secure V2X communication networks for human-centric security and privacy in smart cities. IEEE Access. 12 , 138763–138788 (2024). [ Google Scholar ] 57. Lee, Y. L., Qin, D. & Wang, L. C. 6G massive radio access networks: Key applications, requirements and challenges. IEEE Open. J. Veh. Technol. 2 , 54–66 (2021). [ Google Scholar ] 58. Talebkhah, M. et al. IoT and big data applications in smart cities: Recent advances, challenges, and critical issues. IEEE Access. 9 , 55465–55484 (2021). [ Google Scholar ] 59. Szymanski, T. H. The Cyber Security via Determinism paradigm for a quantum-safe zero-trust deterministic Internet of Things. IEEE Access. 10 , 45893–45930 (2022). [ Google Scholar ] 60. Bhatia, A. S., Saggi, M. K. & Kais, S. Application of quantum-inspired tensor networks to optimize federated learning systems. Quantum Mach. Intell. 7 , 12 (2025). [ Google Scholar ] 61. Kalinin, M. & Krundyshev, V. Security intrusion detection using quantum machine learning techniques. J. Comput. Virol. Hacking Tech. 19 , 125–136 (2023). [ Google Scholar ] 62. Pavithran, D., Shaalan, K., Al-Karaki, J. N. & Gawanmeh, A. Towards building a blockchain framework for IoT. Clust Comput. 23 , 1–15 (2020). [ Google Scholar ] 63. Jurcut, A., Niculcea, T., Ranaweera, P. & Le-Khac, N. A. Security considerations for Internet of Things: A survey. SN Comput. Sci. 1 , 1–19 (2020). [ Google Scholar ] 64. Duong, T. Q., Ansere, J. A., Singh, V. K. & Shin, H. Quantum-inspired machine learning for 6G: Fundamentals, security, resource allocations, challenges, and future research. IEEE Open. J. Veh. Technol. 3 , 375–387 (2022). [ Google Scholar ] 65. Konstantopoulou, E., Athanasiou, G. S., Koutras, N., Kaloxylos, A. & Verikoukis, C. Securing 5G/6G communications in smart cities: Novel SNOW-V/ZUC-256 multimode architectures. IEEE Trans. Netw. Serv. Manag . 20 (2), 2363–2369 (2023). [ Google Scholar ] 66. Bai, Y., Hu, Q. & Seo, S. H. Public participation consortium blockchain for smart city governance. IEEE Internet Things J. 9 (3), 2094–2108 (2022). [ Google Scholar ] 67. DiAdamo, S., Ghibaudi, M. & Cruise, J. Distributed quantum computing and network control for accelerated VQE. Quantum 5 , 597. (2021). 68. Khayyat, M. M. Improved bacterial foraging optimization with deep learning-based anomaly detection in smart cities. Alex Eng. J. 75 , 407–417 (2023). [ Google Scholar ] 69. Khan, M. A., Javaid, S., Hassan, S. A., Kim, B. S. & Kim, S. W. Future-proofing security for UAVs with post-quantum cryptography: A review. IEEE Commun. Mag . 62 (5), 6849–6871 (2024). [ Google Scholar ] 70. Ghorpade, S. N., Zennaro, M. & Chaudhari, B. S. Enhanced differential crossover and quantum particle swarm optimization for IoT applications. IEEE Access. 9 , 93831–93846 (2021). [ Google Scholar ] 71. Tuli, E. A., Lee, J. M. & Kim, D. S. Integration of quantum technologies into metaverse: Applications, potentials, and challenges. IEEE Access. 12 , 29995–30019 (2024). [ Google Scholar ] 72. Shamsoshoara, A., Korenda, A., Afghah, F. & Zeadally, S. A survey on physical unclonable function (PUF)-based security solutions for Internet of Things. Comput. Netw. 183 , 107593 (2020). [ Google Scholar ] 73. Gill, S. S. & Buyya, R. Transforming research with quantum computing. J. Econ. Technol. 1 , 1–15 (2024). [ Google Scholar ] 74. Saez-Ortuño, L., Huertas-Garcia, R., Forgas-Coll, S., Sánchez-García, J. & J Quantum computing for market research. J. Innov. Knowl. 9 (3), 100498 (2024). [ Google Scholar ] 75. Eskandarpour, R., Ghosh, K. J. B. & Aminifar, F. Quantum-enhanced grid of the future: A primer. IEEE Access. 8 , 188993–189002 (2020). [ Google Scholar ] 76. Yang, Z., Alfauri, H., Farkiani, B. & Razmpoosh, R. A survey and comparison of post-quantum and quantum blockchains. IEEE Commun. Surv. Tutor. 26 (2), 967–1002 (2024). [ Google Scholar ] 77. Sheeba, A., Bushra, S. N., Rajarajeswari, S. & Subasini, C. A. An efficient starling-murmuration-based secure web service model for smart city application using DBN. Artif. Intell. Rev. 57 , 1–33 (2024). [ Google Scholar ] 78. Panda, R. C. & Islam, M. S. A deeper look into wind-powered IoT-based sustainable organic compost machine. In: Multimedia Technologies in the Internet of Things Environment , Vol. 2. 25–38. (Springer, 2022). 79. Mahajan, H. B., Rashid, A. S., Junnarkar, A. A. & Uke, N. Integration of Healthcare 4.0 and blockchain into secure cloud-based electronic health records systems. Appl. Nanosci. 13 , 2329–2342 (2023). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 80. Lee, C. C., Tan, T. G., Sharma, V. & Zhou, J. Quantum computing threat modelling on a generic CPS setup. Queen’s University Belfast Technical Report. (2021). 81. Li, X., Wang, C., Li, J. & Zhang, Y. AFED-EF: An energy-efficient virtual machine allocation algorithm for IoT applications in cloud data centers. IEEE Trans. Green. Commun. Netw. 5 (3), 1265–1277 (2021). [ Google Scholar ] 82. Zhang, H., Liu, Y., Xie, S. & Chen, M. ECMS: An edge intelligent energy-efficient model for mobile edge computing. IEEE Trans. Green. Commun. Netw. 5 (4), 1916–1928 (2021). [ Google Scholar ] 83. Kumar, N., Singh, P. & Buyya, R. Energy-efficient task offloading and scheduling for edge–cloud integrated computing systems. Sustain. Comput. : Inf. Syst. 42 , 100972 (2024). [ Google Scholar ] 84. Xu, J., Wang, L. & Chen, G. Cost- and energy-aware service-level scheduling for cloud data centers. J. Supercomput . 75 (9), 5604–5626 (2019). [ Google Scholar ] 85. Anka, F., Gharehchopogh, F. S., Tejani, G. G. & Mousavirad, S. J. Advances in Mountain Gazelle Optimizer: A Comprehensive Study on its Classification and Applications. Int. J. Comput. Intell. Syst. 18 , 247 (2025). [ Google Scholar ] 86. Khoshvaght, P. et al. A multi-objective deep reinforcement learning algorithm for spatio-temporal latency optimization in mobile IoT-enabled edge computing networks. Simul. Model. Pract. Theory . 143 , 103161 (2025). [ Google Scholar ] 87. Sahin, M. F., Yeganli, S. F. & Kiani, F. DSPCI-MTL: Dynamic split point computing in multi-task learning implementation with collaborative intelligence. Alex Eng. J. 124 , 404–421 (2025). [ Google Scholar ] 88. Zhou, Y., Zhang, Y., Wang, X. & Li, K. Deep reinforcement learning for dynamic task scheduling in cloud computing. Computing 105 (8), 1765–1788. 10.1007/s00607-023-01171-z (2023). [ Google Scholar ] 89. Li, H., Zheng, P., Wang, T., Wang, J. & Liu, T. A multi-objective task offloading based on biogeography-based optimization under deadline constraint in mobile edge computing. Cluster Comput. 26 (6), 4051–4067. 10.1007/s10586-022-03809-7 (2023). [ Google Scholar ] 90. Li, H. et al. Energy-efficient offloading based on hybrid bio-inspired algorithm for edge–cloud integrated computation. Sustainable Computing: Inf. Syst. 42 , 100972. 10.1016/j.suscom.2024.100972 (2024). [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement Data is available with the corresponding author and will be given on request. Materials used in this research are available with corresponding author and given on request. The code is available from the corresponding author and can be given on request. Articles from Scientific Reports are provided here courtesy of Nature Publishing Group ACTIONS View on publisher site PDF (7.2 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top