MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641111562 A) filed by Aravind Kumar Karpoorapu; Srilekha Vuyyuru; Meher Deepika Uppaluri; and Kuldeep Chowdary Raavi on September 17, 2026, for An Adaptive Multi-Agent Deep Reinforcement Learning System For Autonomous Cyber Defense.
Inventors include Aravind Kumar Karpoorapu; Srilekha Vuyyuru; Meher Deepika Uppaluri; and Kuldeep Chowdary Raavi.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention discloses the development of an adaptive multi-agent deep reinforcement learning-based system for autonomous cyber defense. The system consists of cybersecurity monitoring modules, a module for constructing the state of the cyber-environment, several autonomous cybersecurity agents, a multi-agent deep reinforcement learning engine, an action coordination module, an action validation module, an action execution module and a feedback-based continuous learning module. The system obtains information about the network, endpoint, identity, application and security events, and creates a dynamic view of the cybersecurity environment. The specialized agents examine various aspects of security and create defensive action suggestions. The multi-agent deep reinforcement learning engine manages the agents and chooses adaptive defense policies based on the threat level, importance of assets, attack phase, effectiveness of the response and its consequences. Validated actions may include endpoint isolation, access limitations, communications interruption, session termination, enhanced monitoring, deception and recovery activities. FIG.1
Disclaimer: Curated by HT Syndication.
