MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641095191 A) filed by Murugalakshmi K; Mr. J. Balaji; Dr. A. Augustus Devarajan; Mrs. M. C. Savithri; and Mr. N. Vijayakumar on August 06, 2026, for Adaptive Cyber Defense Against Phishing Attacks Using Reinforcement Learning.
Inventors include Murugalakshmi K; Mr. J. Balaji; Dr. A. Augustus Devarajan; Mrs. M. C. Savithri; and Mr. N. Vijayakumar.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: Abstract The present invention discloses an adaptive cyber defense system and method for detecting, preventing, and mitigating phishing attacks using Reinforcement Learning (RL). The proposed system employs an intelligent RL agent that continuously interacts with a cybersecurity environment to learn optimal defense policies through reward-based decision-making. The system collects and analyzes multiple security attributes, including email content, URL characteristics, sender reputation, domain information, attachment behavior, network traffic, and user interaction patterns, to identify potential phishing attempts. Unlike conventional phishing detection systems that rely on static rule-based techniques or periodically retrained supervised learning models, the proposed invention dynamically updates its defense strategy by learning from both successful and unsuccessful attack scenarios. The RL agent continuously refines its policy to recognize newly emerging phishing techniques, including zero-day attacks, spear-phishing campaigns, and AI-generated phishing content, without requiring manual intervention. Upon detecting suspicious activities, the system automatically performs adaptive countermeasures such as email quarantine, malicious URL blocking, user alert generation, access restriction, endpoint isolation, and security policy updates. A continuous feedback mechanism enables the RL agent to optimize future responses by maximizing detection accuracy while minimizing false positives and operational overhead. The invention further incorporates threat intelligence integration and real-time monitoring to improve resilience against evolving cyber threats. The disclosed framework provides an autonomous, scalable, and self-learning cybersecurity solution capable of adapting to rapidly changing phishing attack patterns. By continuously improving its detection and response capabilities through reinforcement learning, the invention enhances organizational security, reduces the likelihood of credential theft, financial fraud, and data breaches, and improves the overall effectiveness of enterprise cyber defense systems.
Disclaimer: Curated by HT Syndication.
