SentinelAI

SentinelAI is an autonomous SOC agent that monitors security dashboards, detects rogue AI behavior, and safely disables compromised agents in real time using multimodal analysis and UI navigation.

Comment

Sentinel AI was built to address the emerging security risk of rogue AI agents operating with legitimate system permissions. Inspired by the need for faster-than-human response times in Security Operations Centers (SOCs), Sentinel acts as an autonomous, human-in-the-loop defense agent. It continuously monitors a live SOC dashboard—looking for abnormal API spikes, privilege escalations, and security alerts—and uses Gemini multimodal AI to visually analyze the interface and detect suspicious behavior without relying on fragile, fixed screen coordinates.

Operating on an "Observe, Reason, Decide, Act" loop, Sentinel alerts a human operator when a threat is detected and waits five seconds for intervention. If no response is received, it autonomously navigates the UI to disable the compromised agent. Built with React, Python, and robust fallback mechanisms to handle API outages, Sentinel proves that autonomous agents can safely and reliably defend software environments, paving the way for scalable, AI-driven threat coordination in modern systems.

Built With

Share this project:

Updates