MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641075974 A) filed by Rashmi T V; Madhu Shree R; Jay Prakash Malaviya; Revathi K; Triveni N; Kanchana R; and Divya Shree N on June 19, 2026, for Llm-Powered Autonomous Security Operations Center: Ai-Assisted Threat Intelligence And Automated Incident Response Using Multi- Agent Systems.
Inventor includes Madhu Shree R.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: Modern Security Operations Centers (SOCs) face escalating alert volumes, persistent analyst shortages, and increasingly sophisticated adversaries, resulting in alert fatigue, prolonged dwell times, and inconsistent incident handling. Recent advances in Large Language Models (LLMs) have produced a wave of SOC copilots that assist human analysts, yet these systems remain largely single-agent and human-dependent, providing limited end-to-end automation. This paper proposes an LLM- powered Autonomous SOC built on a multi-agent architecture in which specialized agents collaborate to perform detection, investigation, cyber threat intelligence (CTI) enrichment, decision reasoning, and automated response. A Retrieval-Augmented Generation (RAG) threat-intelligence engine grounds the agents in MITRE ATT&CK techniques, CVE data, and curated CTI reports, while an explainable response-recommendation engine and a human-in-the-loop validation layer preserve analyst oversight and controlled autonomy. We further present a dedicated threat model that treats the LLM-augmented pipeline itself as an attack surface, addressing prompt injection, hallucination, model poisoning, adversarial CTI, and agent manipulation. We hypothesize that coordinated agent collaboration improves alert- prioritization accuracy, reduces analyst workload, and shortens incident-response times, and we outline an evaluation methodology using benchmark intrusion-detection datasets and simulated SOC workflows. The work positions multi-agent autonomous SOC design as a distinct and underexplored research direction relative to existing copilot-centric approaches.
Disclaimer: Curated by HT Syndication.
