MobiLLM: Agentic AI for Automated Threat Mitigation in 6G Open RANs

MobiLLM is a new agentic AI framework designed to automate threat mitigation in 6G Open RAN (O-RAN) architectures. By leveraging a multi-agent system, it moves beyond traditional reactive security to provide an autonomous, closed-loop defense mechanism for next-generation networks.

The primary benefit of MobiLLM is its ability to significantly reduce response latency and manual overhead when defending critical infrastructure. Instead of relying on slow, human-led intervention, the framework uses specialized agents to analyze data, classify threats using trusted sources like MITRE FiGHT and 3GPP, and immediately operationalize countermeasures through O-RAN control interfaces. This ensures that security actions are both precise and rapid enough to match the speed of 6G environments.

To maintain network integrity, MobiLLM incorporates rigorous safety guardrails and human-in-the-loop validation, ensuring that AI-driven decisions remain predictable and trustworthy.

This research was recently presented at the Workshop on 6G and Spectrum Security for Critical Communication (6GSECC), co-located with MILCOM 2025.

You can read the full paper here: https://arxiv.org/abs/2509.21634 or watch this 3-minute introduction video below:

https://www.youtube.com/watch?v=nY9rO1h-SJU