MobiLLM: An agentic AI framework for closed-loop threat mitigation in 6G Open RAN

MobiLLM is an LLM-driven multi-agent framework for fully automated closed-loop threat mitigation in 6G O-RAN environments. It orchestrates security processes through threat analysis, RAG-based threat classification, and threat response agents, executing mitigation actions based on MITRE FiGHT and 3GPP specifications and with the aid of security guardrails. Preliminary evaluation shows that the framework can identify and orchestrate complex mitigation strategies and significantly reduce response latency.