Hybrid ECCMGO-MPCOA framework with ASTRA-based trust and intrusion detection for secure and energy-efficient MANET communication.

Hybrid ECCMGO–MPCOA framework with ASTRA-based trust and intrusion detection for secure and energy-efficient MANET communication - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide 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 Apr 13;16:17208. doi: 10.1038/s41598-026-45323-5 Hybrid ECCMGO–MPCOA framework with ASTRA-based trust and intrusion detection for secure and energy-efficient MANET communication Manojkumar K 1, ✉ , Murugesan S 2 , Suresh Babu A 3 , Sarojini R 4 Author information Article notes Copyright and License information 1 Department of CSE, Government College of Engineering, Thanjavur, India 2 Department of EEE, AMET University, Chennai, India 3 Department of ECE, Karpagam College of Engineering, Coimbatore, Tamilnadu India 4 Department of ECE, Government College of Engineering, bodinayakkanur, Tamilnadu India ✉ Corresponding author. Received 2025 Oct 22; Accepted 2026 Mar 18; 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: PMC13234195  PMID: 41974888 Abstract A mobile ad hoc network (MANET) is a self-configurable network connected by wireless links. This type of network is only suitable for provisional communication links as it is infrastructure-less and there is no centralized control. Wireless Sensor Networks (WSNs) and Mobile Ad Hoc Networks (MANETs) are increasingly employed in mission-critical applications due to the flexibility and scalability. However, the open and dynamic nature of these networks makes it highly susceptible to various security threats, including zero-day attacks, and denial-of-service (DoS), leading to degraded network performance. Mobile Ad Hoc Networks (MANETs) are decentralized and highly dynamic environments that face significant challenges in secure communication and routing efficiency due to node mobility and vulnerability to attacks such as blackhole, Sybil, and DoS. To address these challenges, this framework proposes a novel trust-based secure communication framework that integrates adaptive clustering, optimized routing, and intelligent intrusion detection. The model begins with Density-Aware Possibilistic C-Means (DAPCM) for robust and flexible cluster formation, followed by Enhanced Crisscross Moss Growth Optimization (ECCMGO) for optimal Cluster Head (CH) selection based on energy, trust, and link stability. Each node’s trustworthiness is evaluated through a Multi-Attribute Trust Model (MATM) using behavioral parameters such as packet forwarding ratio, delay, and energy level. These trust scores are embedded into the Modified Pine Cone Optimization Algorithm (MPCOA) to establish secure and efficient routing paths by avoiding low-trust or malicious nodes. To ensure proactive threat detection, the system incorporates an Attention-based Spatio-Temporal Relational Architecture (ASTRA), which captures both spatial interactions and temporal behavioral patterns across nodes. ASTRA enables accurate, real-time detection of attacks by learning complex trust dynamics and network anomalies. Simulation outcomes show significant enhancement in energy efficiency, packet delivery ratio, and intrusion detection accuracy, confirming the model’s suitability for real-time, mission-critical MANET applications. Keywords: Mobile ad hoc network (MANET), Wireless Sensor Networks (WSN), Denial-of-service (DoS), Enhanced Crisscross Moss Growth Optimization (ECCMGO), Cluster Heads (CHs), Multi-Attribute Trust Model (MATM) Subject terms: Engineering, Mathematics and computing Introduction Mobile ad hoc network (MANET) is a wireless mobile network of nodes which is self-forming; that is, a network formed by mobile nodes without relying on an infrastructure. These networks have enough power to support many key applications that would be unfeasible using traditional network infrastructure such as disaster relief, military communications, and sensor networks. The mobile and wireless nature of MANETs poses significant challenges especially power consumption, security and reliability of nodes. Since nodes of MANETs are often battery-powered and often mobile, there is a strong need to maintain adequate communication and provide life to the network 1 . Additionally, the security and reliability of the nodes should be ensured to prevent malicious attacks and untrustworthy behavior that may have a negative impact on the overall performance of the network. Routing protocols can be defined as the means through which routers communicate with each other, so that information sharing and route selection by the nodes in a computer network can occur. The energy efficiency is also a major challenge associated with MANET due to limited power in mobile nodes. Existing routing algorithms do not consider the energy requirements, which is often the cause of premature node failure and reduced network life. The need to have more energy-efficient routing algorithms is seen in high-demand scenarios where nodes receive and send data incessantly. The traditional algorithms can be weak in considering the energy optimisation among the routing decisions, resulting to faster battery loss in nodes and, therefore, the fragmentation of the network 2 . MANETs are increasingly being applied in mission-critical scenarios, such as military coordination, disaster relief and remote sensing that require dynamic, infrastructure-independent communication. Intrinsic characteristics of MANETs including node mobility, lack of centralised control, limited energy sources, etc. makes it exceptionally vulnerable to a wide range of security issues, including black hole attacks, Sybil attacks, and denial of service attacks. Such vulnerabilities may have high impact on integrity, availability, and dependability of data transmission 3 – 5 . Traditional intrusion detection and routing methods do not usually respond dynamically or are not resilient to deal with hostile actions in dynamic topologies. Moreover, current trust-based methods either streamline node behaviour or impose substantial computational expenses. Consequently, there is an immediate necessity for a secure, lightweight, and flexible architecture capable of optimising routing and clustering while simultaneously detecting intrusions in real time with high precision. This work presents an innovative integrated strategy combining a trust model, biomimetic optimisation, and deep learning utilising self-attention, aimed at developing a full security-aware communication system for MANET. The proposed system seeks to harmonise energy efficiency, dependability, and intrusion tolerance, thereby enhancing the security and performance of next-generation self-organising networks 6 . Existing models suggest a game theory-based Fuzzy Secure Clustering (GTFSC) protocol to enhance the security and efficacy of MANETs susceptible to disruption by malevolent nodes. The protocol integrates adaptive evolutionary game theory (EGT), grey wolf optimisation (GWO), and fuzzy clustering for the prediction of threats. Trust values are allocated to nodes and modified according to data transfer behaviour to detect malicious activity. Fuzzy clustering identifies affected nodes, whereas EGT determines the optimal cluster heads for network organisation. GWO creates an effective route from the origin to the destination. The protocol efficiently addresses black hole attacks, grey hole attacks, Sybil attacks, and data manipulation assaults. All identified malicious nodes are eliminated from the transfer pathway to maintain data integrity 7 . The performance findings indicate a packet delivery rate of roughly 98%, a data throughput rate of 90%, a 15% decrease in end-to-end latency, and an 18% reduction in power usage. GTFSC demonstrates exceptional scalability, adaptability, and resilience in both small and large network environments. Another trust-based multi-objective Pelican optimisation algorithm (TMO-POA) enhances safe data transmission in MANETs. Through the integration of encryption, authentication, and trust assessment, the method can proficiently identify and segregate hostile nodes. TMO-POA identifies the most efficient cluster head by evaluating energy efficiency and trust value, hence minimising packet loss and enhancing routing precision. The trust value informs routing choices and guarantees dependable communication, especially in fluctuating and resource-limited settings. This approach tackles significant issues in safe transmission and trust management. Simulation findings indicate that our method surpasses existing methods, including BFOA, C-SSA, and TER. The suggested TMO-POA attains a data throughput of 550.01 Kbps, with an average energy usage of 0.90 J, a packet delivery ratio of 0.999, latency of 11.23 ms, and negligible routing overhead of 0.0550. These findings validate the efficacy, safety, and scalability of MANET setups 8 . The GTFSC model is a good approach based on the fusion of fuzzy clustering with evolutionary game theory and GWO, however, it relies on trust evaluation without changes and does not consider real-time anomaly detection. The rule-based methodologies have limitations in the detection of covert or adaptive attacks and route optimisation is mostly reactive rather than proactive. TMO- POA provides secure channel selection and trust-based routing, however, it does not have a specialised intrusion detection system. The lack of temporal behaviour analysis reduces its effectiveness in handling hybrid and developing threats. Our model to overcome the limitations of existing models uses a Self-attention Encoder Network (SAEN), which is used to detect intrusion in real-time as a sequence-based detector, and it allows the timely identification of burst and progressive attacks. Proactive mitigation of threats can be achieved by the combination of the Multi-Attribute Trust Model (MATM) of assessment of trust dynamics with the PCOA-based trust-aware routing. Moreover, DAPCM-based soft clustering and CCMGO-driven channel selection enhance adaptability and resilience against intricate attacks, including Sybil, DoS, and black hole attacks 9 , 10 . The proposed model incorporates clustering, trust evaluation, optimisation, and intrusion detection to establish a safe and adaptive routing framework for MANETs. In the preliminary phase, clusters are established utilising DAPCM to guarantee adaptability based on the mobility and density of various nodes. ECCMGO is employed to identify the ideal cluster head based on trust score, residual energy, and link stability. The conduct of each node is assessed by a Multi-Attribute Trust Model (MATM), which computes a trust score based on factors including packet forwarding rate, latency, and energy usage. The trust scores inform the MPCOA in selecting secure and energy-efficient routing pathways, thereby circumventing unreliable nodes. From an intrusion detection standpoint, it employs Attention-Based Spatio-Temporal Relation Architecture (ASTRA) to examine spatial linkages and behavioural sequences for the purpose of real-time threat identification. Upon detecting an intrusion, the system modifies trust scores, reconfigures routing, and commences reclustering to ensure safe and dependable connections 11 . The major contribution of the proposed model is as follows: To perform cluster formation, DAPCM is introduced to create flexible and robust clusters that adapt to node density, mobility, and network dynamics. To elect the most suitable cluster heads, ECCMGO is employed to optimize CH selection based on trust score, residual energy, link stability, and node degree. To calculate trust in a dynamic and context-aware manner, a MATM is developed, which evaluates node behavior using parameters like packet forwarding ratio, energy level, delay, and route request patterns. To establish secure and efficient routing paths, the MPCOA is utilized, incorporating real-time trust values to avoid malicious or compromised nodes and minimize energy consumption and delay. To detect network intrusions in real time, a Self-Attention Encoder Network (SAEN) is integrated to analyze sequences of node behavior and accurately identify anomalies such as DoS, Sybil, blackhole, and trust manipulation attacks. To respond effectively to security threats, the system supports dynamic re-routing, trust re-evaluation, cluster head re-election, and cluster reformation when malicious behavior is detected. To demonstrate the effectiveness of the proposed model, extensive simulations are conducted, showing improvements in packet delivery ratio, intrusion detection accuracy, energy efficiency, and communication delay compared to existing approaches. The rest of the framework is structured as follows: Sect. 2 delineates the approach and outcomes of existing models, while Sect. 3 elucidates the operation of the proposed trust-based secure routing protocol, which is coupled with an intrusion detection mechanism and enhanced by a Modified Catch Fish Optimisation Algorithm (MCFOA). The results and comments are presented in Sect. 4 . Concluded the discussion and outlined future work for the suggested system. Literature survey Syeda Mariam Muzammal et al. (2022) proposed the SMTrust model, a secure routing protocol that is based on the idea of trust and mobility metrics, and is aimed at reducing Rank manipulation and Blackhole vulnerabilities in RPL-based Internet of Things. The protocol improves the efficiency of routing by incorporating the mobility-sensitive trust indicators hence ensuring enhanced adaptability in changing IoT environment. The SMTrust is better than MRHOF, the SecTrust, the DCTM, and the MRTS in terms of the packet loss, throughput and topological stability. It has met important routing conditions, such as optimality, consistency and loop-freeness. The technique is not compared to other RPL attacks and might have extra overheads in the very mobile environments 12 . Saad M. Hassan et al. (2024) proposed a trust-aware routing protocol, location-assisted routing protocol, dubbed FLSTMT-LAR, that is based on the LSTM-based trust prediction, federated learning to perform decentralized updates, and multi-objective optimisation to perform routing in MANETs. FLSTMT-LAR enhances the performance of safe and adaptive routing by accurately predicting node behaviour and optimisation of measures of performance like trust, energy efficiency and latency. The protocol delivers packets with a ratio of up to 80% and the network performance is improved by 15% as compared to the traditional single-objective protocols. The model shows robustness in a wide range of network densities, mobility, and malevolent node rates. It can however incur overheads in computation and communication because federated learning and LSTM models in resource-limited nodes are complex 13 . Satheesh Narayanasami et al. (2022) proposed an Enhanced Trust-Based Adaptive Acknowledgement Intrusion-Detection System utilising Multi-objective Particle Swarm Optimisation (MPSO) alongside Kalman filtering to forecast trusted nodes in WSNs. It enhances trust assessment and intrusion detection, optimising data transmission in both secure and hostile contexts. MPSO surpasses regular PSO with a 3.3% enhancement in PDR at 0% harmful nodes and a 3.5% improvement at 30% malicious nodes. Simulations were performed at different levels of network threats to evaluate resilience. The methodology may elevate computational complexity owing to the amalgamation of Kalman filtering with MPSO in resource-constrained sensor nodes 14 . Srivel Ravi et al. (2023) suggested ATSORS-HFO, an Adaptive Trust-based Secure and Optimal Route Selection for MANET, using FBOA, ACGWO, and AES encryption. It guarantees optimum and secure routing by assessing node trust and dynamically selecting K-paths characterised by robust encryption and route stability. Effectiveness is evidenced by assessments across diverse performance indicators under different network conditions. The hybrid optimisation and encryption method may result in heightened computing demands and energy usage on mobile nodes 15 . K. S. Prasanna et al. (2024) introduced a Multiobjective Secure Trust-aware Redundant Array Shifting Encryption with Adaptive Sparrow Search Optimisation (MORS-ASO) for secure data transfer in MANETs. It augments data security by dispersing and dynamically relocating encrypted data between nodes, hence enhancing resistance against routing disruptions and hostile assaults. MORS-ASO attains a packet delivery ratio of 94.3%, exhibits minimal latency of 3.362 s, achieves a high throughput of 2952.75 bps, and sustains a network lifetime of 3500 s. It demonstrates efficient energy consumption (32 J) and operates reliably throughout attack scenarios, sustaining 19% packet loss and 1.245 ms jitter. The amalgamation of several optimisation and encryption methodologies may elevate algorithmic complexity and processing overhead for low-resource MANET devices 16 . Deepika Kukreja et al. (2022) designed a Trust-based Secure and Energy-Aware (T-SEA) routing protocol for the identification and exclusion of black/gray hole nodes in MANETs by selective intrusion detection. It enhances network longevity and security by designating IDS duties to nodes that are highly trusted, energy-efficient, and well-connected, eliminating the necessity for all nodes to engage in sniffing. T-SEA facilitates precise identification of malicious nodes while minimising energy use by activating the Intrusion identification System alone when required. The methodology is validated by NS-2 simulations, illustrating its efficacy in detecting and mitigating hazards. The performance of T-SEA may deteriorate in high-mobility circumstances or if IDS nodes are degraded or lack energy 17 . Mohan Das et al. (2024) suggested the trust-based Self-Detection Routing (TSDR) protocol to be used with a Stable Intrusion Detection System (S-IDS). The combined scheme ensures high level of route maintenance, accuracy of trust as well as effective identification and isolation of malicious nodes, which improves the resilience of the network. S-IDS+TSDR outperforms protocols like AODV, GTM and ESCT by wide margins in empirical measurement of factors like packet delivery ratio, throughput, delivery probability and latency in varying scenarios. The protocol is robust to dynamically changing and hostile environment and is also scalable and provides secure communication; performance decreases slightly with greater density of malicious nodes and performance decreases more with higher node density 18 . Sripriya et al. (2022) proposed a trust-based Quality of Service (QoS) routing protocol in MANETs that combines social and QoS trust. The protocol enhances forwarding node selection and judges its usefulness based on parameters like residual energy, the quality of the channel, and the reliability of the connection and thus increases QoS and security. It is effective in identifying and isolating bad or idle nodes, and reducing inaccuracies in packet collection, and improving stability in routing. Simulations that are run using NS 2 are more precise and effective in ensuring secure and high-quality pathways in MANETs. The model can have challenges in environments of mobile connections or fast-changing networks because it relies on the computation of trust and behavioral measurements 19 . Meena Rao et al. (2023) presented a trust-based safe routing protocol using Stacked Autoencoder with Sparse Bayesian Regression (SAE-SBR) to forecast trust values and PCK-means to categorise nodes. Despite 10 rogue nodes, the technique achieves 99.5% packet delivery, 94% detection accuracy, and 0.023s end-to-end delay. The technique maximises routing by avoiding low-trust nodes, improving dependability and performance (95.568\x3841.522mbps). The simulation findings show that secure routing can be considerably enhanced under hostile attack. Regression-based trust evaluation may have scalability concerns in large networks, dynamic topologies, or several attack strategies 20 . Shalini Subramani et al. (2024) presented a Fuzzy-ant colony optimisation secured quality of service routing protocol (FEDSSRP) with a unique Intrusion Detection System (IDS) for safe and energy-efficient WSN multihop routing. QoS and trust measures help the clustering algorithm locate cluster heads and the IDS detects dangerous activity. F-ACO-SQoSRP optimises and secures route selection, improving packet delivery ratio, network life, latency, and energy utilisation. The simulation shows that the suggested model outperforms secure routing methods in trust-based security, communication reliability, and energy usage. Fuzzy logic, trust, and clustering algorithm might raise computing needs, especially for sensor nodes with restricted resource 21 . Rahma Meddeb et al. (2023) developed a Stacked Autoencoder-based Intrusion Detection System (Stacked-AE-IDS) to secure MANET. To improve intrusion detection for denial-of-service attacks, the Stacked AE-IDS minimises feature correlation and develops high-level features. A Deep Neural Network detector (DNN-IDS) receives the autoencoder’s encoding to improve the system’s DoS attack detection based on MANET routing service factors. This method improves detection and IDS resilience. Quality and availability of annotated datasets may limit model efficacy. Deep architecture training (Stacked AE + DNN) can be costly for real-time implementation in resource-constrained MANET environments 22 . Huangshui Hu et al. (2021) proposed a model named as TBSEER which uses adaptive direct trust, indirect trust from the Sink node, and energy trust to appropriately analyse node behaviour. This integrated approach lets TBSEER detect and defend against black hole, selective forwarding, sinkhole, hello flood, and wormhole routing threats. To speed up malicious node detection, an adaptive penalty mechanism and volatilisation factor are used. By outsourcing indirect trust computation to the Sink, the protocol conserves node energy. Simulation results show that TBSEER improves security, energy efficiency, and network lifespan. The Sink’s indirect trust computations may be a single point of failure, and the protocol may have scalability and latency issues in highly dynamic or large-scale installations 23 . Nikitina Vlada et al. (2023) propose the EPSO-IDSRA, an Enhanced Particle Swarm Optimization-based Intrusion Detection and Secure Routing Algorithm for MANETs, using fuzzy clustering and confidence-based node evaluation for optimal routing.The model is safe, energy efficient, with a 0.008ms delay, 0.15 ms energy usage, 0.8bps throughput, and 89% intrusion detection. Nevertheless, confidence metrics and cost of fuzzy clustering can result in poor performance of the system in the highly dynamic or in the large-scale network environment 24 . Saleh A. Alghamdi et al. (2021) constructed a Trust-Aware Intrusion Detection and Prevention System (TA -IDPS) in 5G cloud-integrated MANETs through ultra-lightweight symmetric cryptography, Moth Flame Optimisation, Deep Belief Networks, and cloudlet-based packet validation. TA-IDPS authenticates and classifies mobile nodes, fraudulent packets detection, energy saving, scaling, and better results than the current methods are its detection accuracy, delay, throughput, and overhead. Categorisation and cloudlet-layer entropy analysis based on deep-learning in large and real-time deployment can be computationally costly because the model can be complex 25 . Neenavath Veeraiah et al. (2022) introduced a safe multipath routing system for MANETs utilising the Bird Swarm-Whale Optimisation Algorithm (BSWOA), integrating fuzzy clustering for cluster head selection and fuzzy Naive Bayes for intrusion detection. The model enhances energy efficiency, security, and routing performance for assaults such as flooding and blackhole, with significant energy retention (9.48 J), throughput (0.676 bps), detection rate (69.9%), and minimal delay (0.00372 ms). Although performance improvements are evident, the hybrid optimisation paradigm may present computational complexity and scalability issues in extensive or highly dynamic MANET contexts 26 . Alagan Ramasamy Rajeswari et al. (2022) developed ANFIS-based Energy-Efficient Secure Clustering model (ANFIS -EESC) of MANETs, including a Weight-Based Trust Estimation (WBTE) algorithm and a Fuzzy-Based Clustering (FBC) algorithm. The model improves on network durability (by 39%), packet delivery rate (by 22%−32%), and throughput (by 32%−39%) in comparison to the existing Weighted Clustering Algorithm (WCA) and it is an effective solution to rogue nodes. The combination of ANFIS and other fuzzy logic techniques can increase the requirements of the computational process, which in turn can hamper real-time adaptation in dynamic or resource-constrained MANET settings 27 . Ujwala Ravale et al. (2024) proposed a Game Theory-based Fuzzy Secure Clustering (GTFSC) protocol for Mobile Ad Hoc Networks (MANETs), utilising Evolutionary Game Theory (EGT) for optimal clustering, Fuzzy Clustering for the detection of malicious nodes, and the Grey Wolf Optimisation (GWO) algorithm for efficient routing. The model achieves exceptional performance with a 98% packet delivery ratio, 90% throughput, a 15% reduction in end-to-end delay, and an 18% drop in energy consumption, while effectively identifying and removing rogue nodes. The amalgamation of diverse algorithms (EGT, GWO, fuzzy logic) may lead to increased system complexity and computational load, potentially affecting scalability and responsiveness in highly dynamic environments 28 . Table 1 summarises the existing models. Table 1. Existing work Summary table. Author Name & Year Proposed Model Outcome Limitation Syeda Mariam Muzammal (2022) SMTrust: Trust and mobility-based secure routing protocol for IoT Defends against RPL Rank and Blackhole attacks; improves packet loss, throughput, and topology stability Not tested against other RPL attacks; may introduce overhead in highly mobile scenarios Saad M. Hassan (2024) FLSTMT-LAR: Trust-aware and location-aided routing with LSTM and federated learning 80% PDR; improves network performance by 15%; robust under varying densities and mobility patterns Computational and communication overhead due to LSTM and federated learning in resource-constrained nodes Satheesh Narayanasami (2022) Enhanced Trust-Based Adaptive Acknowledgment IDS using MPSO and Kalman filter Improves trust evaluation and intrusion detection; 3.5% PDR gain at 30% malicious nodes Increases computational complexity due to MPSO and Kalman filtering Srivel Ravi (2023) ATSORS-HFO: Adaptive Trust-based Secure Routing using FBOA and ACGWO with AES encryption Improves throughput, PDR, detection rate; reduces latency and packet loss Hybrid optimization and encryption increase computational load and energy consumption K. S. Prasanna (2024) MORS-ASO: Secure routing with Redundant Array Shifting Encryption and Adaptive Sparrow Search 94.3% PDR; low delay (3.362s); high throughput (2952.75 bps); 3500 s network lifetime Algorithmic complexity and processing overhead in low-resource MANET devices Deepika Kukreja (2022) T-SEA: Trust-based Secure and Energy-Aware routing protocol with selective IDS Accurate detection of black/gray hole nodes; reduces energy consumption by activating IDS only when needed Performance may decline in high mobility or if IDS nodes are compromised or energy-depleted Mohan Das (2024) TSDR + S-IDS: Trust-based Self-Detection Routing with Intrusion Detection System Outperforms AODV, GTM, ESCT in PDR, throughput, latency; robust under attacks and mobility Slight performance degradation with increased malicious nodes; higher jitter in dense networks G. Sripriya (2022) Trust-based QoS Routing Protocol integrating social and QoS trust Improves secure communication and reliable forwarding; effective malicious/dead node isolation May face issues in high-mobility or dynamic networks due to trust computation reliance Meena Rao (2023) SAE-SBR with PCK-means: Trust-based secure routing with autoencoder and Bayesian regression 99.5% PDR; 94% detection accuracy; low delay (0.023s); avoids low-trust nodes for optimal routing Regression-based trust evaluation may not scale well in highly dynamic or large-scale networks Shalini Subramani (2024) F-ACO-SQoSRP: Fuzzy-based Ant Colony Optimization Secure QoS Routing with IDS Enhances packet delivery, network lifetime, and energy efficiency; secure and reliable multi-hop routing Fuzzy logic, clustering, and IDS integration increase computational overhead, challenging for resource-constrained sensor nodes Open in a new tab Proposed methodology The proposed hybrid ECCMGO–MPCOA framework integrates adaptive clustering, trust-aware optimization, and attention-based intrusion detection into a unified MANET security architecture. Density-Aware Probabilistic C-Means (DAPCM) enables density-adaptive cluster formation, while ECCMGO performs multi-criteria cluster head selection by jointly optimizing residual energy, trust, mobility, and link stability. A dynamic Multi-Attribute Trust Model continuously evaluates node behavior, and the computed trust scores are embedded into MPCOA to construct secure, energy-efficient routing paths by excluding malicious or unstable nodes. To enhance proactive defense, a Self-Attention Encoder Network (SAEN) captures temporal trust variations and anomalous communication patterns for real-time intrusion identification. This tightly coupled optimization–trust–detection strategy ensures resilient, energy-balanced, and attack-aware communication in highly dynamic MANET environments 29 . Figure 1 shows the block diagram of proposed Intrusion Detection. Fig. 1. Open in a new tab Block diagram of proposed Intrusion Detection for Secure MANET Communication model. Cluster formation using density-aware possibilistic C-means (DAPCM) Density-Aware Possibilistic C-Means (DAPCM) is adopted in the proposed framework to ensure stable and adaptive cluster formation under highly dynamic MANET conditions. Traditional clustering approaches often suffer from frequent re-clustering and instability due to node mobility and uneven node distribution. To address this limitation, DAPCM integrates possibilistic membership modeling with local density awareness, enabling robust soft clustering that reduces sensitivity to noise and mobility fluctuations. By assigning density-weighted membership values, nodes in dense regions exert stronger influence during centroid updates, resulting in balanced and resilient clusters. This mechanism significantly minimizes cluster head turnover and control overhead, thereby improving energy conservation and network lifetime. The primary purpose of employing DAPCM is to establish a stable structural foundation for subsequent trust evaluation and secure routing. By maintaining cluster consistency even in sparse or rapidly changing topologies, DAPCM enhances communication reliability and supports the efficient execution of ECCMGO-based cluster head selection and trust-aware routing optimization within the proposed secure MANET architecture 30 . Cluster head selection using enhanced crisscross moss growth optimization (ECCMGO) Traditional CCMGO Traditional Cross-Moss Growth Optimization (CCMGO) is employed for cluster head (CH) selection following DAPCM-based cluster formation. The algorithm models CH selection as a multi-objective optimization problem, where candidate nodes are evaluated using residual energy, node degree, mobility factor, link stability, and trust score. These parameters collectively ensure that the selected CH maintains connectivity, energy balance, and communication reliability within the cluster. CCMGO enhances the conventional cross-search strategy by introducing trust-threshold filtering, preventing nodes with low behavioral credibility from being considered for leadership. The search mechanism iteratively explores candidate solutions in alternating directions to identify an optimal CH based on aggregated fitness evaluation. Compared to single-parameter clustering methods, CCMGO incorporates both performance and security-related metrics, enabling more stable and resilient cluster leadership in dynamic MANET environments. The multi-criteria formulation improves robustness under mobility variations and reduces the probability of selecting compromised or unstable nodes as cluster heads. Cross-Colour Moss Growth Optimisation (CCMGO) is a naturally inspired metaheuristic algorithm that emulates the adaptive growth behaviour of mosses in the natural habitats. Mosses in nature have a cross- or multi-directional growth pattern, expanding both horizontally and vertically towards regions with optimal water, fertiliser, and light availability. This tendency is essential to CCMGO’s capacity to investigate and utilise the search space of an optimisation problem 31 . Crisscross strategy The CC strategy, derived from the CSO algorithm, incorporates Horizontal Crossover Search (HCS) and Vertical Crossover Search (VCS) to enhance optimisation performance. HCS supports information sharing among different individuals, whereas VCS promotes interaction across multiple dimensions within a single individual. This dual crossover mechanism, combined with competitive selection, improves global exploration and accelerates convergence. By integrating CC into MGO, the algorithm achieves better population diversity and avoids premature convergence to local optimal. Horizontal crossover search The HCS mechanism randomly selects particles from the population and pairs them to perform crossover operations. By incorporating demographic information, it enhances exploration capability. Its mathematical representation is provided in Eqs. ( 1 ) and ( 2 ). 1 2 Here, and represent random numbers in [0, 1], and indicate random variables in [− 1, 1], and and indicate the jth dimension of the ith and kth particles, respectively. and are the new HCS offspring from the two particles. HCS then uses greedy selection to keep the fittest offspring and progenitors. Vertical crossover search The VCS process randomly selects dimensions within each particle and couples them for crossover execution. By leveraging individual-level information, it strengthens exploitation capability. The detailed mathematical expression is presented in Eq. ( 3 ). 3 r3 is a random number between [0, 1], xj1pi and xj2i are the values of the two dimensions chosen at random by the ith individual, and reflects the VCS-produced offspring. VCS, like HCS, uses greedy selection to preserve high-fitness children and ancestors. Dynamic population divisions parameter To address the scalability concern, the population division parameter in MGO is redesigned as a dynamic function of iteration number and population size , rather than using the fixed value . In high node density scenarios (large ), the division factor automatically adapts to maintain balanced exploration and exploitation. During early iterations, fewer divisions allow broader population influence to enhance global search, while later iterations increase divisions to refine local convergence. This adaptive mechanism ensures stable direction updates and consistent optimisation performance even as network size and problem dimensionality increase. 4 Where FEs is the current number of evaluations, is the maximum, and dim is the issue dimension. The proposed ECCMGO The proposed Enhanced Crisscross Moss Growth Optimization (ECCMGO) is designed for secure and energy-aware Cluster Head (CH) selection in dynamic MANET environments. Conventional CH selection approaches rely primarily on single metrics such as residual energy or node degree, making it vulnerable to malicious node selection and instability under mobility. To address these limitations, ECCMGO formulates CH selection as a multi-objective optimization problem integrating trust score, residual energy, node degree, mobility factor, and link stability. Trust Integration : In contrast to traditional CCMGO, which emphasises exploration efficiency via inter-population information exchange and adaptive search direction, ECCMGO incorporates a trust score as a pivotal parameter in the optimisation process. The trust ratings are calculated using a Multi-Attribute Trust Model (MATM), which assesses node behaviour based on packet forwarding consistency, energy consumption, latency, and anomaly patterns. This integration guarantees that malicious or errant nodes are systematically eliminated from consideration as Cluster Head (CH) candidates, irrespective of the energy or connectivity characteristics. By integrating trust into the optimisation framework, ECCMGO improves the reliability and security of the clustering procedure. Multi-Objective Fitness Function The original CCMGO predominantly assesses candidates through standardised performance indicators. ECCMGO substitutes this with a multi-objective fitness function specifically designed for MANETs. The function incorporates five parameters: trust score ( ), residual energy ( ), node degree ( ), mobility ( ), and connection stability ( ). The employed scoring formula is: 5 The coefficients in Eq. ( 5 ) are selected to reflect the relative importance of security, energy efficiency, connectivity, mobility, and link stability in cluster head selection. All parameters are normalized within [0,1], and the weights satisfy to ensure balanced contribution. Higher weights are assigned to trust and residual energy to prevent malicious or low-energy nodes from being elected as cluster heads. Sensitivity analysis shows that increasing the trust weight improves security metrics, while increasing the energy weight enhances network lifetime. The chosen coefficient configuration provides an optimal trade-off between stability, security, and overall network performance. Trust Threshold Filtering To enhance security, ECCMGO implements a trust threshold technique. Nodes having a trust score within a specified threshold are immediately eliminated from the optimisation pool, irrespective of any advantageous characteristics. This modification explicitly precludes the selection of hostile or compromised nodes as Cluster Heads, an aspect not addressed in the original CCMGO architecture. Context-Aware Re-election Strategy CCMGO has a fixed iteration-based update mechanism, whereas ECCMGO improves adaptability through event-triggered cluster head re-election. This transpires in reaction to immediate alterations, such as a decline in trust score or notifications from the intrusion detection system. This enables the model to consistently preserve secure and efficient communication, even in unstable or adversarial network environments. The specified adjustments convert CCMGO from a general-purpose optimiser into a security-conscious, trust-oriented, and MANET-specific cluster head selection algorithm, hence augmenting its relevance in practical wireless network contexts.As illustrated in Fig. 2 , ECCMGO employs iterative cross-search exploration with controlled diversity mechanisms to avoid premature convergence while maintaining solution robustness. Fig. 2. Open in a new tab Flowchart of proposed ECCMGO. Open in a new tab Algorithm 1 : Pseudo-code of the ECCMGO Trust-aware routing process The suggested structure introduces a trust-sensitive routing algorithm that incorporates dynamically evaluated trust appraisals and Modified Pine Cone Optimization (MPCOA) as a way of guaranteeing safe and energy-efficient data transfer within MANETs. The routing fitness function includes node trust scores as calculated by the Multi-Attribute Trust Model (MATM) along with the latency and energy values. The nodes which have a trust score lower than a set threshold are not included in the path to avoid malicious commenting in routing. When the trust is compromised or a threat is identified, the optimiser automatically reconstructs paths to maintain resilient and attack-aware communication in conditions of dynamic networks 32 . Trust evaluation using multi-attribute trust model (MATM) The suggested model applies a MATM to guarantee the credible and behaviour-conscious decision-making in the dynamic MANET settings. In contrast to single-metric trust schemes, MATM uses a dynamic trust score, which combines packet-forwarding ratio, remaining energy, delay, frequency of route-request, and anomaly. This multi-dimensional assessment allows effective discrimination between collaborative and malicious nodes and even fits the behavioural changes by time. The calculated trust scores have a direct impact on cluster-head selection, routing optimisation, so that only those nodes which behave in a reliable manner are included in the important communication processes. The inclusion of trust evaluation into the network core improves the security, stability and robustness against internal attacks associated with MATM. In the suggested MATM network, a watchdog node W oversees the action of assessed node B per a number of predetermined trust qualities . Monitoring and gathering behavioural evidence is done by the evaluating node or a specified neighbour. The assessing node, in the second stage, combines its own observation and the one reported by watchdog nodes of each trust attribute. Lastly, the total value of trust of the tested node is calculated as a weighted summation of the trust attributes. The three-phase approach facilitates proper, collaborative and dependable trust assessment. Trust attributes : Trust qualities are the elements that influence trust levels, represented by ρ. Each trust feature value varies between 0 and 1. Prior to examining the specifics of our trust program in MANETs, it will first evaluate the fundamental trust attributes and subsequently outline our trust model. It has delineated the subsequent trust attributes for managing control and data traffic within the proposed trust framework. Control packet generation : A control packet is the protocol-specific data exchanged between nodes to establish routes and sustain topology. The trust property enables an assessing node to evaluate the reliability of the assessed node about control packet generating behaviour as defined by the routing protocol. The results of node W concerning node B about control packet creation are articulated by the subsequent Eq. 6 While t represents the current time, p represents the actual number of control messages produced by node B during the interval (t, t + Ω) as perceived by W, and represents the expected number of messages. An assessing node later integrates its observations with those reported by the watchdog nodes to ascertain a reputation for node B, as seen in the following Eq. 7 where α represents the weight factor assigned to the observations of an evaluating node and those of a watchdog node. Control packet forwarding : Nodes in a MANET depend on mutual cooperation to relay traffic. A non-cooperative forwarding node may either delete packets or forward control packets with delays, resulting in a distorted understanding of the network structure. It will designate the packets that are successfully overheard as “pack.” The evaluations of node W regarding node B in relation to control packet forwarding can be determined using the subsequent Eq. 8 An evaluating/watchdog node W has a minimum packet loss rate of 0 and a maximum of 1, indicating that misbehaving nodes trash all sent packets. An assessing node then combines its findings with those of watchdog nodes to determine node B’s control packet forwarding repute. 9 In the equation, α is the weight factor allocated to data from the assessing and watchdog nodes. Data packet forwarding : Nodes control traffic and relay data packets. A node may discard and retransmit data packets with latency or updated information. The following equation calculates node W’s data packet forwarding assessments of node B: 10 where ξ signifies the aggregate quantity of data packets transferred and pack indicates the data packets successfully identified by the watchdog node W. By synthesising the assessments from the evaluating node and the watchdog node, it ascertain the cumulative reputation of the evaluated node for data packet forwarding, as articulated in the subsequent Eq. 11 Open in a new tab Algorithm 2 : Multi-Attribute Trust Model (MATM) Trust Evaluation Algorithm Routing using modified pine cone optimization algorithm (MPCOA) The algorithm, which is used to obtain secure, energy-conscious and adaptive routing in dynamic MANET conditions, is the Modified Pine Cone Optimization Algorithm (MPCOA) that becomes susceptible to maliciousness in the routing of malicious nodes due to the traditional shortest-path strategies. The choice of MPCA is based on the fact that it has a good global-search and quick convergence, and hence it is able to be used in the multi-objective route optimisation in mobility and adversarial conditions. Candidate routes in the proposed framework are evaluated using a composite fitness function that integrates the average trust score derived from MATM, cumulative end-to-end delay, and residual energy. This multi-objective formulation enables balanced route selection by simultaneously maximizing trustworthiness while minimizing delay and energy consumption. The nodes that have a lower trust value than a specified trust level are blocked before route construction to avoid malicious activities. Recursive route exploration This spiral-based exploration algorithm recursively narrows route suggestions in the trust-compliant search area until an excellent secure path is found, which provides resilient and performance-sensitive and attack-conscious routing in dynamic settings in a MANET 33 , 34 . Mathematical modeling of pine reproduction The Pine Cone Optimization Algorithm (PCOA) is inspired by the natural processes of pollination and cone seed dispersal in pine trees. It is used to solve complex optimization problems by mimicking swarm intelligence and evolutionary behavior observed in nature. The purpose of this algorithm is to efficiently explore the search space and identify the global optimal solution while avoiding local optima. In the process, pollen particles represent search agents that move within the search space similar to wind-driven pollination. The transfer of pollen between cones models exploration, while cone development represents solution improvement. Seed dispersal by gravity and fauna imitates population updating and diversification. If a better position (solution) is found, it replaces the previous one, ensuring convergence toward the global optimum. Generating the initial population in PCOA The PCOA consists of two populations: pine trees and the cones. The issue domain is divided into equal pieces. During the method initialisation, a tree is situated at the center of each segment, encircled by numerous generated cones. The boundary specifications for each component are delineated as follow: 12 13 In Eqs. ( 12 ) and ( 13 ), lb and ub signify the lower and upper bounds of the problem’s domain, whereas UbS and LbS express the upper and lower bounds of each segment, accordingly. The population of pine cones is generated in the following method. 14 In Eq. ( 14 ), CX indicates the position of the pine cone, designates a random vector conforming to a normal distribution between 0 and 1, and rand represents a random number likewise conforming to a normal distribution within the same range. In the operators of PCOA, i signifies the pine tree index and j indicates the cone index. Upon setting up PCOA, the location of each tree aligns with the position of the ideal cone within that segment. Exploitation The exploitation phase enhances solutions locally via gravity-based and animal dispersal-inspired mechanisms. Gravity-driven local search emulates the descent of a cone influenced by gravitational forces: 15 Animal dispersal induces adaptive perturbations to solutions: 16 The weight decay factor progressively amplifies exploitation intensity over time. Exploration Phase During the exploration phase, the system utilises a pollination-based update, influenced by wind currents and interactions among cones. The update for the candidate solution is provided by: 17 In Eq. ( 17 ), ϕ represents the probability of successful pine cone pollination. The value of ϕ is determined using the following formulation: 18 19 In Eqs. ( 18 ) and ( 19 ), signifies the effect of the ith cone on the jth cone, while indicates the Euclidean distance between the ith and jth cones. Moreover, α, β, and γ are invariant parameters that formally define the pollination process. Proposed MPCOA The MPCOA augments the original PCOA to include the trust-awareness and security requirements making it suitable to routing in a secure environment in MANETs. The original PCOA focuses on the path optimisation via the spiral dispersion of seeds to balance exploration and exploitation via position update and convergence to an optimal solution, however, it lacks a node trustworthiness assessment mechanism, which is necessary in the insider threat prone environment such as blackhole, Sybil and grayhole attacks. Trust-Integrated Cost Function : In contrast to the original PCOA, which typically downplays measures such as distance or energy, MPCOA implements a multi-objective cost function that assesses possible paths according to: Average trust score of the nodes in the path. Total end-to-end delay. Average residual energy. The modified cost function is defined as: 20 Where denotes average trust score of all nodes in the path, represents the total delay from source to destination represents the average residual energy of the path λ1,λ2,λ3 are weight coefficients such that . The weight coefficients in Eq. ( 20 ) are assigned to balance security, delay, and energy efficiency in routing decisions, with the constraint . All parameters are normalized to ensure fair contribution within the cost function. A higher emphasizes secure, high-trust paths, while prioritizes low-latency routes and promotes energy-balanced transmission. Sensitivity analysis shows that adjusting these weights directly influences PDR, delay, and network lifetime. The selected configuration ensures an optimal trade-off between reliability, performance, and sustainability in dynamic MANET environments. Intrusion detection using ASTRA The proposed model incorporates ASTRA to provide intelligent and lightweight intrusion detection in highly dynamic MANET environments. ASTRA is employed to overcome the limitations of conventional threshold-based and watchdog mechanisms, which often fail to detect coordinated, slow, or stealthy attacks. Its purpose is to accurately identify malicious behaviours such as blackhole, grayhole, Sybil, and DoS attacks while maintaining low computational overhead and reduced false positives. Sequential trust metrics, packet-forwarding behaviour, route-request patterns, energy variations, and inter-node relatedness are the metrics to be analysed by the detection process. Using an attention mechanism, ASTRA simultaneously measures temporal behavioural changes and spatial dependencies between nodes, and thus can detect anomalies according to the context. The suspiciousness of activity is detected, and the node is decreased in its trust score and is not used in the routing (MPCOA) and cluster-head selection processes. This combination guarantees the real time, adaptive and robust security enforcement within the MANET structure 35 , 36 . Attention-based spatiotemporal relational architecture (ASTRA) The model proposes ASTRA as a lightweight and intelligent intrusion detection that will work in very dynamic MANET environment. ASTRA is used to counter the shortcomings of traditional threshold-based and watchdog systems that are usually not able to detect slow, coordinated, or stealthy attacks. It aims at the ability to precisely detect malicious behaviours like blackhole, grayhole, Sybil, and DoS attacks at low computational cost and low false positive. Within the detection algorithm, the activity of each node is denoted by a sequence of spatiotemporal features that consist of variations of trust, forwarding behaviour at packets, frequency of route requests, links stability, and energy usage. The features are fed through a multi-head self-attention mechanism wherein adaptive weights of importance are given to key behavioural and inter-node patterns. When an abnormal activity is detected, ASTRA decreases the trust value of a node, using MATM, and informs the routing (MPCOA) and clustering modules to isolate the compromised node. It is a real-time, built-in mechanism that can guarantee scalable, context-sensitive, and low-false-positive intrusion-detection in highly dynamic MANET environments 36 . Result and discussions The proposed framework was tested in terms of performance and compared to the current solutions of MANET, including HMRFOA, TMO-POA, RDO-ENB, ATSORS-HFO, and HO-TESRO. The aim of such an evaluation was to confirm the suitability of trust modelling, optimisation, and intelligent intrusion detection when used collectively under one architecture. Key metrics such as the ratio of packet delivery, end-to-end delay, energy and intrusion-detection accuracy were used to analyse performance. The proposed model delivered better reliability in terms of packet delivery by eliminating low-trust and malicious nodes, and adaptive re-routing minimized latency in communication in dynamic situations. Balance in the selection of cluster-heads and controlled use of resources made the energy efficiency better. Moreover, ASTRA integration has increased the accuracy of detection and minimised false positives using spatiotemporal patterns of behaviour. All in all, the findings indicate that the trust-based and optimisation-based framework provides better security, stability and performance over the current MANET routing techniques 37 – 39 . As can be seen in Table 2 , the simulation experiments were carried out in a 100 m x 100 m MANET environment with 500 randomly distributed nodes. The network is of IEEE 802.11b standard with 2Mbps channel capacity and a transmission distance of 250 m. The Nodes are based on the Random Waypoint mobility model and the speeds range between 1 and 20 m/s and the pause time ranges a period of 0–30 s. CBR traffic over UDP with 25 connections is created with 1024-byte packets and each node is charged with 102 J of energy. The simulation time is 1000 s and the same is run on several occasions with varying fixed random seed (e.g., 25,50,75) to ensure statistical reliability. Table 2. Simulation parameter setup. Parameters Values Total number of nodes 500 Network size 100 m × 100 m Wireless standard IEEE 802.11b Channel capacity 2 Mbps Packet size 1024 bytes Initial energy 10 J Node placement Random Mobility model Random Waypoint Node speed 1–20 m/s Pause time 0–30 s Simulation time 1000 s Traffic type CBR over UDP Number of connections 25 Transmission range 250 m Random seed Fixed seed (e.g., 25, 50, 75 for multiple runs) Carrier frequency 2.4 GHz Open in a new tab Energy consumption The total energy (in mJ) expended by network nodes during communication and routing processes is referred to as energy consumption. Table 3 ; Fig. 3 demonstrate that the energy consumption of all approaches correlates with the number of nodes due to increased communication overhead. Nonetheless, the suggested model exhibits the lowest energy usage across all network sizes, recording 0.58 mJ for 100 nodes, 0.76 mJ for 200 nodes, 0.92 mJ for 300 nodes, 0.96 mJ for 400 nodes, and 1.01 mJ for 500 nodes. The suggested framework will conserve a substantial amount of energy compared to existing techniques, such as HO-TESRO and RDO-ENB, owing to its balanced clustering and trust-aware routing mechanisms. This demonstrates its efficacy in prolonging the network’s lifespan and optimising resource utilisation. Table 3. Performance Comparison of Energy Consumption and Throughput Metrics. Algorithms Energy Consumption (mJ) Throughput (Mbps) No.of nodes 100 200 300 400 500 100 200 300 400 500 ATSORS-HFO 1.29 1.4 1.52 1.59 1.82 0.83 0.75 0.72 0.67 0.60 HMRFOA 1.24 1.31 1.43 1.48 1.76 0.92 0.89 0.86 0.77 0.74 TMO-POA 0.97 1.08 1.28 1.44 1.63 0.94 0.91 0.83 0.82 0.73 RDO-ENB 0.82 1.02 1.09 1.25 1.31 0.95 0.93 0.87 0.86 0.79 HO-TESRO 0.75 0.95 1.03 1.12 1.25 0.97 0.95 0.92 0.91 0.87 Proposed model 0.58 0.76 0.92 0.96 1.01 0.99 0.96 0.95 0.94 0.93 Open in a new tab Fig. 3. Open in a new tab Performance evaluation of Energy consumption. Throughput Throughput is the effective speed of data transmission, measured in Mbps, within a network during a specified duration. Most of the existing methods have been shown to gradually reduce the throughput with the increase in node density as a result of congestion and malicious interference. As indicated in Table 3 ; Fig. 4 , the proposed model demonstrates very high values of throughput that are at 0.99 Mbps (100 nodes), 0.96 Mbps (200 nodes), 0.95 Mbps (300 nodes), 0.94 Mbps (400 nodes), and 0.93 Mbps (500 nodes). Secure routing (MPCOA) and real-time intrusion detection (ASTRA) are used to maintain stability in throughput by averting malicious disruptions. These findings prove that the model can maintain stable data transmission even in thick MANETs environment. Fig. 4. Open in a new tab Performance evaluation of Throughput. End to end delay (E2ED) (s) E2ED is an average duration (in seconds) it takes a packet of data to pass the network between the source and the destination. Table 4 ; Fig. 5 show that delay tends to worsen with the number of nodes in the current techniques because it is congested with the route not being well maintained. Nonetheless, the delayed values of the proposed model are much lower (2.5s in 100 nodes, 3.0s in 200 nodes, 3.2s in 300 nodes, 3.6s in 400 nodes, and 4.01s in 500 nodes). This loss is reduced by trust-based optimal routing as well as quick reconfiguration in case of malicious node detection. The findings prove that the suggested framework makes communication faster and more stable than the current methods. Table 4. Performance Comparison of E2ED and Network lifetime Metrics. Algorithms E2ED (s) Network lifetime (rounds) No.of nodes 100 200 300 400 500 100 200 300 400 500 ATSORS-HFO 6.3 6.6 7.3 7.9 8.2 3850 3650 3450 3250 3150 HMRFOA 6.1 6.3 6.7 7.6 7.4 4250 3850 3650 3450 3300 TMO-POA 5.4 5.7 6.1 6.4 7.1 4600 4250 4100 3900 3600 RDO-ENB 5.1 5.4 5.7 6.1 6.6 4850 4600 4350 4350 4200 HO-TESRO 4.6 5.1 5.3 5.9 6.3 5200 4950 4400 4600 4150 Proposed model 2.5 3.0 3.2 3.6 4.01 5450 5400 5100 5000 4000 Open in a new tab Fig. 5. Open in a new tab Performance evaluation of E2ED. Network lifetime (rounds) Network lifetime is the total time of operation (number of rounds) when a big fraction of nodes exhaustthe energy. In the majority of conventional techniques, the network lifetime reduces with the density of nodes (because of the unequal energy usage, and overloading of the cluster heads). As indicated in Table 4 ; Fig. 6 , this model has the ability to realize a long lifetime of 5450 rounds (100 nodes), 5400 rounds (200 nodes), 5100 rounds (300 rounds), 5000 rounds (400 nodes), and 4000 rounds (500 nodes). This enhancement has been credited to balanced selection of cluster heads and routing schemes that are energy conscious. These findings suggest that the suggested strategy leads to an increase in the sustainability of the entire network and long-term stability. Fig. 6. Open in a new tab Performance evaluation of Network lifetime. PDR (%) PDR is the ratio of delivered data packets that manage to reach the destination. In the current approaches, PDR is declining with the node density slowly because of congestion, interference and malicious node involvement. Table 5 ; Fig. 7 show that the proposed model has better PDR values of 98.7% (100 nodes) 98.0% (200 nodes), 97.0% (300 nodes), 95.8% (400 nodes), and 94.0% (500 nodes). This advancement can be explained by the trust-sensitive routing and real-time intrusion detecting, which exclude the unreliable nodes during communication. These findings indicate an improved data reliability and network strength as the network scale is increased. Table 5. Performance Comparison of PDR and PLR Metrics. Algorithms PDR (%) PLR (%) No.of nodes 100 200 300 400 500 100 200 300 400 500 ATSORS-HFO 90 87.8 87.2 85 82 11 12.1 13.3 14.5 15.8 HMRFOA 93 91.3 88.9 87.3 85 10 10.5 11.2 12 14.2 TMO-POA 94.5 93.7 91.8 90.6 88 9 9.6 10.1 10.3 11.5 RDO-ENB 95.2 94.5 94 92.5 90 6 7.9 8.8 9 10.2 HO-TESRO 96.4 96.4 94.9 93.7 92 5 6.5 7.4 7.6 9.1 Proposed model 98.7 98 97 95.8 94 3 3.8 5.1 5.9 6.4 Open in a new tab Fig. 7. Open in a new tab Performance evaluation of PDR. PLR(%) PLR refers to the proportion of the data packets which do not get to the destination when it is sent across. In the traditional strategies, the larger the network, the larger is the PLR caused by unstable routes and security susceptibility. Table 5 , and Fig. 8 indicate a significant reduction in the packet loss, which is 3%, 3.8%, 5.1%, 5.9%, and 6.4% (100 nodes, 200 nodes, 300 nodes, 400 nodes and 500 nodes respectively). This is done by reducing path security and isolating malicious nodes fast. The reduced PLR proves that the framework is effective in ensuring that the data transmission is in a stable and safe environment. Fig. 8. Open in a new tab Performance evaluation of PLR. Jitter (ms) Jitter is defined as the deviation of the time span in which packets arrive (in milliseconds) when transmitting data. A larger jitter means that communication is not stable and that the quality of service is lower, particularly in real-time applications. Table 6 ; Fig. 9 demonstrate that with increment in node density there is a steady increase in jitter of existing methods as a result of congestion and routing instability. Conversely, the proposed model shows considerably lesser values of jitter 28ms (100 nodes), 33ms (200 nodes), 35ms (300 nodes), 38ms (400 nodes), and 40ms (500 nodes). The decrease is obtained by the means of stable trust-based routing and quick threat isolation, which guarantee the smoother and more stable delivery of data. Table 6. Performance Comparison of Jitter and Bit error rate Metrics. Algorithms Jitter (ms) Bit error rate (%) No.of nodes 100 200 300 400 500 100 200 300 400 500 ATSORS-HFO 53 56 59 61 63 23 24 21.5 19.5 17 HMRFOA 46 50 54 59 60 21 20 19.4 18.4 16 TMO-POA 43 47 50 54 57 19 17.5 16 15 14 RDO-ENB 40 44 47 50 52 17 16 14 13 12 HO-TESRO 36 38 44 46 49 13 12 11 10 9 Proposed model 28 33 35 38 40 9 8 7.5 7 5.8 Open in a new tab Fig. 9. Open in a new tab Performance evaluation of jitter. Bit error rate (%) BER is a ratio of the number of bits that are received in the wrong order because of interference, noise or unstable routing paths. The traditional methods have relatively elevated error rate with increase in network size. Table 6 ; Fig. 10 are used to show that the proposed model has the lowest values of BER of 9% (100 nodes), 8% (200 nodes), 7.5% (300 nodes), 7% (400 nodes), and 5.8% (500 nodes). This is enhanced by safety in the choice of the path, malicious interference is minimized and the stability of the link is enhanced. The reduced BER verifies the improved transmission accuracy and reliability of communication in the suggested framework in general. Fig. 10. Open in a new tab Performance evaluation of BER. Efficiency rate (%) It is a measure of the overall performance effectiveness of the network as regard to success in resource utilization and dependable data transmission. Increase in efficiency implies an increased coordination among the clustering, routing and security mechanisms. As it can be seen in Table 7 ; Fig. 11 , although the existing methods demonstrate a gradual decrease in performance with an increase in node density, the proposed model demonstrates high efficiency values of 99.82%, 99.70%, 99.35%, 99.12%, and 99.08%. The enhancement is brought about by combined trust-sensitive routing, equal clustering, and smart intrusion detection. Such findings support the strength of the suggested scheme and its streamlined functionality. Table 7. Performance Comparison of Efficiency rate and routing overhead ratio Metrics. Algorithms Efficiency rate (%) routing overhead ratio No.of nodes 100 200 300 400 500 100 200 300 400 500 ATSORS-HFO 97.9 97.9 97.42 96.7 95.3 0.34 0.36 0.41 0.43 0.47 HMRFOA 98.2 98.2 97.67 97.2 96.4 0.28 0.29 0.36 0.38 0.43 TMO-POA 98.4 98.4 98.19 97.5 97.1 0.22 0.26 0.33 0.37 0.39 RDO-ENB 98.7 98.6 98.23 98.3 97.6 0.19 0.24 0.29 0.32 0.36 HO-TESRO 99.6 99.4 98.89 98.8 98.4 0.16 0.21 0.24 0.27 0.34 Proposed model 99.82 99.7 99.35 99.12 99.08 0.12 0.17 0.19 0.25 0.29 Open in a new tab Fig. 11. Open in a new tab Performance evaluation of efficiency rate. Routing overhead ratio It refers to a %age of control packets sent in route establishment and maintenance in comparison to real data packets sent. Decreased overhead means that there are more efficient routing and less control traffic. Table 7 ; Fig. 12 indicate that the proposed model displays the lowest values of overheads of 0.12 (100 nodes), 0.17 (200 nodes), 0.19 (300 nodes), 0.25 (400 nodes), and 0.29 (500 nodes), relative to the current strategies. This is brought about by secure path validation, less re-transmissions and adaptive route maintenance. The reduced routing overhead proves to be better in scaling and efficiency of communication in dense MANET networks. Fig. 12. Open in a new tab Performance evaluation of routing overhead. Attack detection rate ADR (%) It is the ratio of malicious actions, which are recognized well by the system of intrusion detection. The greater the values, the greater the ability of identifying the direct and coordinated attacks. As shown in Table 8 ; Fig. 13 , the detection rates of the existing methods are moderate, and with an increase in the density of the nodes, the detection rates increase linearly, however, only to up to 95%, 97%, 97.9%, and 98.7% with the increase in the node density of the proposed model. This advancement can be credited to the attention based spatiotemporal analysis that has been carried out by ASTRA. The outcomes confirm that the suggested method provides a very accurate and reliable identification of threat in dynamic MANET conditions. Table 8. Performance Comparison of ADR and security Metrics. Algorithms ADR (%) Security analysis (sec) No.of nodes 100 200 300 400 500 100 200 300 400 500 ATSORS-HFO 75 75.7 79 81 82.5 0.82 0.99 1.07 1.64 1.92 HMRFOA 76.4 78.3 79.7 84 86 0.98 1.46 1.61 1.79 2.16 TMO-POA 77 80.1 83.6 88 89.5 1.12 1.58 1.69 1.91 2.65 RDO-ENB 79 80.9 85.2 89 90 1.63 1.73 1.75 1.98 2.83 HO-TESRO 80.5 84.7 88.3 92 94 2.2 2.34 2.53 2.63 2.94 Proposed model 95 96.5 97.4 97.9 98.7 2.7 2.92 3.36 3.57 4.36 Open in a new tab Fig. 13. Open in a new tab Performance evaluation of ADR. Security performance (s) It is the amount of time that the system needs to analyse network behaviour and identify possible intrusions. Even though sophisticated detection systems might need a little more processing time, it is more accurate and stronger. Table 8 ; Fig. 14 indicate that analysis times of the proposed model are 2.7s (100 nodes), 2.92s (200 nodes), 3.36s (300 nodes), 3.57s (400 nodes), and 4.36s (500 nodes). The processing time is scalable even though it increases with the size of the network. This illustrates that the proposed intrusion detection mechanism is highly accurate in detection and at the same time, the mechanism is realistic with regard to real-time performance. Fig. 14. Open in a new tab Performance evaluation of security analysis. Memory consumption Memory consumption is the size of the storage units used by the routing and security mechanisms in the course of network operation. The reduced memory usage is a sign of lightweight and scalable frameworks that can be used in resource-limited MANET environments. Table 9 ; Fig. 15 indicate that existing methods have relatively large memory consumption with increase in node density but the proposed model has lower memory consumption of 640 bytes (100 nodes), 670 bytes (200 nodes), 690 bytes (300 nodes), 700 bytes (400 nodes), and 720 bytes (500 nodes). The minimisation is realised by optimised computation of trust, clustering with efficacy and lightweight design of intrusion detection. These findings affirm the fact that the suggested framework operates at a high level without the need to spend a lot on memory. Table 9. Performance Comparison of Memory consumption and Computational time Metrics. Algorithms Memory consumption (bytes) Computational time (ms) No.of nodes 100 200 300 400 500 100 200 300 400 500 ATSORS-HFO 760 810 795 820 815 3.3 3.6 4.7 5.2 5.8 HMRFOA 730 740 755 775 790 2.5 2.9 4.2 4.6 5.4 TMO-POA 700 725 740 760 780 2 2.6 3.7 4.3 5.1 RDO-ENB 690 710 715 720 750 1.9 2 3.5 4.1 4.7 HO-TESRO 670 690 700 710 735 1.7 1.8 3.2 3.8 4.1 Proposed model 640 670 690 700 720 1.3 1.5 2.6 3.4 3.6 Open in a new tab Fig. 15. Open in a new tab Performance evaluation of memory consumption. Computational time (ms) Computational time refers to the duration required for executing routing decisions, clustering processes, and security analyses. Processing latency must be reduced to enable real-time communication through effective algorithms. Table 9 ; Fig. 16 demonstrate that the suggested model exhibits reduced computation times of 1.3 ms (100 nodes), 1.5 ms (200 nodes), 2.6 ms (300 nodes), 3.4 ms (400 nodes), and 3.6 ms (500 nodes) in comparison to existing methodologies. This is enhanced by the optimal selection of routes, equitable distribution of cluster heads, and streamlined intrusion detection processing. The reduced computation time demonstrates that the model is applicable in large-scale and dynamic MANETs. Fig. 16. Open in a new tab Performance evaluation of computational time. Statistical analysis Table 10 suggests that the statistical approach of 500 nodes proves the Proposed Model to be the most successful in terms of the PDR (94.0%) and throughput (0.93 Mbps) to be the most successful method compared to all the methods. The low values of the standard deviation are used to show consistent and stability in performance in more than one simulation run. The 95 30% confidence intervals indicate the obvious non-overlapping of the ranges with the baseline methods and ensure the statistical reliability. The proposed approach also enjoys high scalability and resilience even when the node density is high. The findings confirm the model efficacy in massively-scaled MANETs. Table 10. Statistical Significance Analysis of Performance Metrics. Algorithms PDR (%) Throughput (Mbps) (Mean ± SD) 95% CI (Mean ± SD) 95% CI ATSORS-HFO 82.0 ± 0.40 [81.0, 83.0] 0.60 ± 0.03 [0.57, 0.63] HMRFOA 85.0 ± 0.38 [84.1, 85.9] 0.74 ± 0.04 [0.69, 0.79] TMO-POA 88.0 ± 0.45 [86.9, 89.1] 0.73 ± 0.03 [0.69, 0.77] RDO-ENB 90.0 ± 0.42 [89.0, 91.0] 0.79 ± 0.03 [0.75, 0.83] HO-TESRO 92.0 ± 0.40 [91.0, 93.0] 0.87 ± 0.02 [0.84, 0.90] Proposed Model 94.0 ± 0.35 [93.2 , 94.8] 0.93 ± 0.02 [0.90 , 0.96] Open in a new tab Computational complexity Table 11illustrates that the computational complexity of the proposed model is O(N²) + O(I⋅P⋅D) + O(H⋅T⋅F), reflecting trust evaluation, multi-objective optimisation, and attention-based intrusion detection, respectively. The O(N²) term arises from pairwise trust and relational node analysis, ensuring reliable node validation. The O(I⋅P⋅D) component represents the iterative optimisation process in MPCOA. The O(H⋅T⋅F) term corresponds to ASTRA’s spatiotemporal attention processing. Despite integrating security and optimisation modules, the overall complexity grows in a controlled and modular manner, ensuring scalability and practical feasibility. Table 11. Computational Complexity Evaluation. Algorithms Computational complexity ATSORS-HFO HMRFOA TMO-POA RDO-ENB HO-TESRO Proposed model Open in a new tab State of the art comparison Table 12 ; Fig. 17 indicate that the suggested DAPCM-ECCMGO-MPCOA model offers the highest accuracy of 99.89%, recall of 98.63% recall, and 96.42% TPR, which are better than any of the latest approaches. Even though Ujwala Ravale et al. (2024) mentioned high accuracy (98.03%) with EGT-GWO-fuzzy clustering, and Suga Priya et al. (2025) mentioned high recall (97.95%) with RL-CHS + PUF+ALEA, all of these models lack overall precision and consistency. Additional methods like FuzzyNB-BSWOA, EPSOIDRA and MSSA have moderate performances but it is unreliable in TPR. Conversely, the suggested system provides high-accuracy, balanced intrusion detection and high-quality threat response based on all metrics. Table 12. State-of-the-art comparison. Reference no Author name Algorithm Accuracy (%) Recall (%) TPR (%) 26 Neenavath Veeraiah et al. (2022) Fuzzy NB-BSWOA 90.14 92.2 87,3 24 Nikitina Vlada et al. (2023) EPSO-IDSRA 88.5 90.65 84.72 28 Ujwala Ravale et al. (2024) EGT, GWO, and fuzzy clustering 98.03 97.38 95.34 35 Lakshminarayana et al.(2023) TMO-POA 81.47 95.24 85.6 35 S. Suga Priya et al. (2025) RL-CHS + PUF (QR-PUF)+ALEA 80.32 97.95 89.5 38 Layth Hussein et al. (2024) MSSA 84.96 96.57 86 Proposed DAPCM – ECCMGO – MPCOA 99.89 98.63 96.42 Open in a new tab Fig. 17. Open in a new tab Performance evaluation of ADR. Table 13 ; Fig. 18 indicate that the proposed framework of DAPCM-ECCMGO-MPCOA provides the most ADR of 98.56% Instead, the lowest energy consumption is 1.01 mJ, and it is better than all the recent models. Although Layth Hussein et al. (2024) and Suga Priya et al. (2025) had 93% and 85.6% detection respectively with MSSA and RL-CHS + PUF+ALEA, both of it used more energy. Other systems like TMOPOA and EPSOIDSRA had fair detection rates (81% and 77.52%) at a higher cost of energy. The suggested model guarantees a greater level of qualified threat detection and a reduced use of energy thus it is suitable in resource-limited MANET setting. Table 13. Comparative Analysis of ADR and Energy Consumption. Reference no Author name Algorithm ADR (%) Energy consumption 26 Neenavath Veeraiah et al. (2022) Fuzzy NB-BSWOA 74.7% 2.02 mJ 24 Nikitina Vlada et al. (2023) EPSO-IDSRA 77.52% 1.82 mJ 28 Ujwala Ravale et al. (2024) EGT, GWO, and fuzzy clustering 79.3% 1.61 mJ 35 Lakshminarayana et al.(2023) TMO-POA 81% 1.34 mJ 35 S. Suga Priya et al. (2025) RL-CHS + PUF (QR-PUF)+ALEA 85.6% 1.22Mj 38 Layth Hussein et al. (2024) MSSA 93% 1.19 mJ Proposed DAPCM – ECCMGO - MPCOA 98.56% 1.01mJ Open in a new tab Fig. 18. Open in a new tab Performance evaluation of Energy consumption. Discussion Mechanistic justification of performance improvements This efficient functionality of proposed framework lies in the coordinated response of trust modelling (MATM), energy-conscious clustering (ECCMGO), multi-objective routing (MPCOA), and attention-based intrusion detection (ASTRA). As compared to the current schemes that optimise individual metrics (e.g., hop count or residual energy), the above architecture combines trust validation with optimisation, hence only trusted nodes are involved in communication. The given trust-conscious filtering is a direct reason as to why the Packet Delivery Ratio (PDR) and the Packet Loss Ratio (PLR) have been improved as the malicious or volatile nodes are eliminated before route construction. The composite fitness function in MPCOA improves throughput and decreases End-to-End Delay (E2ED) as minimises delay and cost of energy at the same time maximises trust. As routing decisions have been pre-tested with behavioural trust scores, the occurrence of route failures and re-discovery process is minimized causing a decrease in jitter and a better transmission stability. Moreover, ECCMGO allocates the duties of cluster heads in terms of remaining energy and trustworthiness, and node overloading is avoided, which increases the life of the network and decreases the total energy usage. The high rate of attack detection can be explained by the fact that ASTRA employs the spatial-temporal attention mechanism which retrieves both temporal behavioural anomalies and inter-node pattern of relations. This allows coordinated and stealthy attacks to be detected early minimizing disruption propagation within the network. Initial separation of compromised nodes reduces the routing overhead and elimination of redundant exchange of control packets. Furthermore, there are lightweight representations of trust and selective search-space reduction in MPCOA and ECCMGO that explain the reduced memory consumption and computation time. In general, the gains in the performance are not independent gains but results of architectural synergy: trust-based filtering leads to more stable routing, energy-based clustering leads to more sustainable performance, and attention-based intrusion detection leads to the maintenance of security integrity. This combined design guarantees scalable, robust and resource efficient MANET functioning in both dynamic and adversarial situations. Scalability considerations for higher node densities Although the experimental evaluation in this study is conducted up to 500 nodes, the scalability of the proposed Hybrid ECCMGO–MPCOA framework can be analytically justified based on its architectural design and computational characteristics. In conventional MANET routing schemes, performance degradation at high node densities occurs primarily due to global route discovery overhead, uncontrolled broadcast flooding, and centralized trust evaluation mechanisms. In contrast, the proposed framework adopts a distributed and cluster-based architecture, which inherently limits scalability bottlenecks. First, the Density-Aware Possibilistic C-Means (DAPCM) algorithm partitions the network into bounded clusters, ensuring that control operations are localized within cluster regions rather than executed network-wide. This confines clustering complexity to O(nk)O(nk)O(nk), where kkk represents the number of clusters, preventing exponential growth in overhead as node density increases. Second, Enhanced Crisscross Moss Growth Optimization (ECCMGO) performs Cluster Head selection within individual clusters. Since optimization is restricted to local node subsets, the search space remains limited even as the overall network size grows. Third, the trust-aware Modified Pine Cone Optimization Algorithm (MPCOA) restricts route discovery to high-trust nodes, effectively reducing redundant retransmissions and broadcast storms—two dominant causes of scalability degradation in dense MANETs. Finally, the ASTRA-based intrusion detection mechanism operates on localized spatio-temporal node interactions rather than requiring global topology awareness. Its inference complexity scales proportionally to active communication links, enabling computational feasibility in higher-density deployments. Therefore, although empirical validation beyond 500 nodes is reserved for future work, the distributed clustering strategy, localized optimization processes, and trust-pruned routing collectively indicate that the proposed framework is structurally scalable and avoids centralized bottlenecks typical of traditional MANET security protocols. Advantages The proposed model provides an integrated and smart approach towards safe and effective communication in MANETs with the combination of trust-sensitive routing, self-governing clustering, and current intrusion detection. To support a broad network resilience, this paradigm uses many layers of decisions making, unlike with normal protocols that optimise routing or clustering separately. Cluster Head selection (enhanced Crisscross Moss Growth Optimisation or ECCMGO) is an energy-based prioritisation that values the connection and behavioural trustworthiness of the nodes, thus ensuring that compromised or unstable nodes are not given the important task to perform. The integration of the ASTRA module, which is based on attention-based spatiotemporal learning, gives a powerful defensive shell, which allows the system to foresee and isolate rogue nodes before it affects communication. This significantly increases the reliability of packet delivery, reduces loss, as well as, decreasing the latency. The model also applies trust thresholding and selective re- routing in order to reduce unnecessary computations and overheads in communications and maintain low complexity levels regardless of the stratified architecture. It also ensures that there is fairness in energy consumption thus extending the network life. The combination of the above factors makes the presented architecture highly responsive to the mobility of nodes and adjusting to dynamic patterns of attacks, and at the same time, it is scalable to both small-scale and large-scale systems. The architecture of the model is cohesive and context-sensitive that makes it more efficient and secure than the contemporary trust-based or optimisation-based approaches. Conclusion The proposed approach effectively tackles the primary difficulties of safe and efficient communication in MANETs by incorporating dynamic clustering, trust-aware routing, and intelligent intrusion detection into a cohesive framework. The system guarantees safe and energy-efficient data transmission in highly dynamic and hostile settings by employing Density-Aware Possibilistic C-Means (DAPCM) for adaptive clustering, ECCMGO for secure Cluster Head selection, and MPCOA for trust-based routing. The incorporation of the ASTRA facilitates strong real-time intrusion detection through the analysis of spatiotemporal behaviour patterns. The simulation findings indicate that the proposed model regularly surpasses contemporary state-of-the-art methods across various essential performance measures, including packet delivery ratio, end-to-end delay, energy consumption, network longevity, throughput, and intrusion detection accuracy. The model guarantees that the routing overhead is reduced, the memory used is minimized, as well as the complexity of computations, and, therefore, the model becomes more feasible when applied to real-time and resource-constrained MANET applications. The proposed system is a combination of trust evaluation, adaptive optimisation, and context-sensitive intrusion detection that offers a strong, scalable, and secure mobile network communication infrastructure in the next generation. Future research may concentrate on expanding this approach to heterogeneous networks, incorporating blockchain for decentralized trust management, or enhancing the IDS for reduced latency and improved scalability in ultra-dense network environments. Author contributions The first author should be that person who contributed most to the work, including writing of the manuscript. The sequence of authors should be determined by the relative overall contributions to the manuscript. Funding This research received no external funding. Data availability All data analyzed during this study are included in this article. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. Srilakshmi, U., Alghamdi, S. A., Vuyyuru, V. A., Veeraiah, N. & Alotaibi, Y. A secure optimization routing algorithm for mobile ad hoc networks. IEEE Access 10 , 14260–14269 (2022). [ Google Scholar ] 2. Kumar, A. et al. Advanced network security with an integrated trust-based intrusion detection system for routing protocol. Alex. Eng. J. 120 , 378–390 (2025). [ Google Scholar ] 3. Danjuma Maiwada, U., Danyaro, K. U. & Aminu Abdulkadir Aliyu. Aliza Bt Sarlan, &. Dynamic Handover Optimization Protocol to enhance energy efficiency within the A-LTE 5G network’s two-tier architecture. International Journal of Data Informatics and Intelligent Computing, 3(3), 8–15. (2024). 4. Sankaran, K. S. & Hong, S. P. Trust aware secured data transmission based routing strategy using optimal ch selection in mobile ad-hoc network. Mob. Netw. Appl. 29 (2), 385–397 (2024). [ Google Scholar ] 5. Sharma, S. & Hussain, S. Z. A survey of trust based secure routing protocol used in mobile ad hoc networks. In ITM Web of Conferences (Vol. 54, p. 02009). EDP Sciences. (2023). 6. Lakshmi, G. V. & Vaishnavi, P. A trusted security approach to detect and isolate routing attacks in mobile ad hoc networks. J. Eng. Res. 12 (3), 379–386 (2024). [ Google Scholar ] 7. Alyoubi, A. A. Enhancing data security in mobile ad-hoc network (MANETs) using trust-based approach with RSSI and fuzzy logic. Mob. Networks Appl. 29 (6), 2030–2046 (2024). 8. Mahamune, A. A. & Chandane, M. M. Trust-based co-operative routing for secure communication in mobile ad hoc networks. Digit. Commun. Netw. 10 (4), 1079–1087 (2024). [ Google Scholar ] 9. Isah, S., Agushaka, J. O. & Agber, S. Trust-Aware Energy Optimization Techniques in Mobile Ad-hoc Networks: Current Trends and Future Directions. In Proceedings of the Faculty of Science Conferences (Vol. 1, pp. 65–79). (2024). 10. Phan, T. C. & Tran, H. C. Consideration of data security and privacy using machine learning techniques. International Journal of Data Informatics and Intelligent Computing 2 (4), 20–32 (2023). [ Google Scholar ] 11. Veerabadrappa, K. & Lingareddy, S. C. Secure routing using multi-objective trust aware hybrid optimization for wireless sensor networks. Int. J. Intell. Eng. Syst , 15 (1), 540–548 (2022). 12. Muzammal, S. M., Murugesan, R. K., Jhanjhi, N. Z., Hossain, M. S. & Yassine, A. Trust and mobility-based protocol for secure routing in Internet of Things. Sensors 22 (16), 6215 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Hassan, S. M., Mohamad, M. M., Muchtar, F. B. & Dawoodi, F. B. Y. P. Enhancing MANET Security through Federated Learning and Multiobjective optimization: A Trust-aware Routing Framework (IEEE Access, 2024). 14. Narayanasami, S., Butta, R., Govindaraj, R., Choudhary, S. S., Sharma, D. K., Poonia,A., … Verma, R. (2022). An enhanced trust-based Kalman filter route optimization technique for wireless sensor networks. Wireless Personal Communications, 1–19. 15. Ravi, S., Matheswaran, S., Perumal, U., Sivakumar, S. & Palvadi, S. K. Adaptive trust-based secure and optimal route selection algorithm for MANET using hybrid fuzzy optimization. Peer-to-Peer Netw. Appl. 16 (1), 22–34 (2023). [ Google Scholar ] 16. Prasanna, K. S. & Ramesh, B. Multiobjective secure trust aware redundant array shifting encryption and clustering based routing in mobile ad hoc networks. Int. J. Commun. Syst. 38 (5), e6074 (2025). [ Google Scholar ] 17. Kukreja, D. & Sharma, D. K. T-SEA: trust based secure and energy aware routing protocol for mobile ad hoc networks. Int. J. Inform. Technol. 14 (2), 915–929 (2022). 18. Oliver, A. S., Reddy, G. V., & Sarumathi, S. Integrating trust-based self-detection and stable intrusion detection for scalable and secure routing in mobile ad hoc networks. 10.2139/ssrn.4995114. 19. Sripriya, G. & Santha, T. A trust-based design for secure and quality of service routing in mobile ad hoc networks. Int. J. Comput. Netw. Appl. 9 (5), 522. (2022). [ Google Scholar ] 20. Rao, M., Chaudhary, P., Sheoran, K. & Dhand, G. A secure routing protocol using hybrid deep regression based trust evaluation and clustering for mobile ad-hoc network. Peer-to-Peer Netw. Appl. 16 (6), 2794–2810 (2023). [ Google Scholar ] 21. Subramani, S. & Selvi, M. Intrusion detection system and fuzzy ant colony optimization based secured routing in wireless sensor networks. Soft. Comput. 28 (17), 10345–10367 (2024). [ Google Scholar ] 22. Meddeb, R., Jemili, F., Triki, B. & Korbaa, O. A deep learning-based intrusion detection approach for mobile ad-hoc network. Soft. Comput. 27 (14), 9425–9439 (2023). [ Google Scholar ] 23. Hu, H., Han, Y., Yao, M. & Song, X. Trust based secure and energy efficient routing protocol for wireless sensor networks. IEEE Access 10 , 10585–10596 (2021). [ Google Scholar ] 24. Vlada, N. et al. Enhancing security in mobile ad hoc networks: enhanced particle swarm optimization-driven intrusion detection and secure routing algorithm, J. Wirel. Mob. Networks Ubiquitous Comput. Dependable Appl. 14 (3), 77–88 (2023). 25. Alghamdi, S. A. Novel trust-aware intrusion detection and prevention system for 5G MANET–Cloud. Int. J. Inf. Secur. 21 (3), 469–488 (2022). [ Google Scholar ] 26. Veeraiah, N. & Krishna, B. T. An approach for optimal-secure multi-path routing and intrusion detection in MANET 1–15 (Evolutionary Intelligence, 2022). 27. Rajeswari, A. R., Lai, W. C., Kavitha, C., Balasubramanian, P. K. & Srividhya, S. R. A trust-based secure neuro fuzzy clustering technique for mobile ad hoc networks. Electronics 12 (2), 274 (2023). [ Google Scholar ] 28. Ravale, U. & Borkar, G. M. Secure and efficient trust enabled routing in mobile ad hoc network using game theory and grey wolf optimisation techniques. IET Wirel. Sens. Syst. 14 (6), 451–476 (2024). [ Google Scholar ] 29. Singh, A., Raj, A., Rani, P., Khatibi, A., Aldeeb, H., Shukla, P. K., … Hassan, M.M. (2025). Resilient wireless sensor networks in industrial contexts via energy-efficient optimization and trust-based secure routing. Peer-to-Peer Networking and Applications,18(3), 1–17. 30. Kumar, V. S. A big data analytical framework for intrusion detection based on novel elephant herding optimized finite Dirichlet mixture models. International Journal of Data Informatics and Intelligent Computing 2 (2), 11–20 (2023). [ Google Scholar ] 31. Yue, T. & Li, T. Crisscross moss growth optimization: An enhanced bio-inspired algorithm for global production and optimization. Biomimetics 10 (1), 32 (2025). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Aravindan, S. & Rajaram, A. Energy-aware multi-attribute trust modal for secure MANET-IoT environment. Multimedia Tools Appl. 83 (38), 85637–85662 (2024). 33. Valikhan Anaraki, M. & Farzin, S. The pine cone optimization algorithm (PCOA). Biomimetics 9 (2), 91 (2024). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Kumuthini, C., Thangarasu, N., Kavitha, K. & Gomathy, K. Ant with artificial bee colony techniques in vehicular ad-hoc networks. International Journal of Data Informatics and Intelligent Computing 2 (3), 21–28 (2023). [ Google Scholar ] 35. Priya, S. S., Vijayabhasker, R. & Rajaram, A. Advanced Security and Efficiency Framework for Mobile Ad-Hoc Networks Using Adaptive Clustering and Optimization Techniques 1–12 (Journal of Electrical Engineering & Technology, 2025). 36. Hamza, F. & &Vigila, S. M. C. A trust management scheme for intrusion detection system in MANET using weighted Naïve Bayes classifier, Int. J. Eng. Trends Technol. 70 (2), 75–85 (2022). 37. Ilakkiya, N. & Rajaram, A. A secured trusted routing using the structure of a novel directed acyclic graph-blockchain in mobile ad hoc network internet of things environment. Multimedia Tools Appl. 83 (40), 87903–87928 (2024). 38. Tr, M. & Kumar, V. V. Clustering techniques for recommendation of movies. International Journal of Data Informatics and Intelligent Computing 1 (2), 16–22 (2022). [ Google Scholar ] 39. Gracy Theresa, W., Prakash, M. & Betina Antony, J. Multicast on-route cluster propagation using to identify the network intrusion detection system in mobile ad hoc network. Int. J. Commun Syst. 34 (11), e4850 (2021). Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement All data analyzed during this study are included in this article. Articles from Scientific Reports are provided here courtesy of Nature Publishing Group ACTIONS View on publisher site PDF (4.7 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases 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

Related documents

• Hybrid intelligence-powered secure clustering with trust-optimized routing for next-generation MANET communication.#123016
• ZK-eSIM: A Privacy-Centric Zero-Knowledge Approach for eSIM Provisioning#667924
• Survey: federated learning data security and privacy-preserving in edge-Internet of Things#341900
• An enhanced ECC for securing data, comparative study.#274858
• Evasion Attacks on Cost-Utility-Based Adversarial Training for Online AutoML in IoT Networks#1115937