An LLM-Based Agentic Network Traffic Incident-Report Approach Towards Explainable-AI Network Defense

Traditional intrusion detection systems for IoT networks achieve high classification accuracy but lack interpretability and actionable incident-response capabilities, limiting their operational value in security-critical environments. This paper presents a graph-based multi-agent framework that inte