MUMBAI, India, Jan. 2 -- Intellectual Property India has published a patent application (202541122994 A) filed by Vellore Institute Of Technology, Vellore, Tamil Nadu, on Dec. 5, 2025, for 'radar framework for adversarially robust cyber defense ai agents.'
Inventor(s) include Dr. S. Dilipkumar; Mr. Manoj Chandrasekaran; and Ms. Tanshya Mishra.
The application for the patent was published on Jan. 2, under issue no. 01/2026.
According to the abstract released by the Intellectual Property India: "The present disclosure provides a computer-implemented method for generating adversarial examples to evade cyber defense AI agents. The method implements a reparameterized Variational Actor-Critic process comprising an actor network and a critic network (5). The method applies a Gumbel-Softmax transformation to enable differentiable sampling from discrete action spaces, wherein the Gumbel-Softmax transformation converts discrete action selection into a continuous differentiable process using a temperature parameter. The method generates adversarial examples by learning a policy for selecting sequences of discrete adversarial actions against a target cyber defense AI agent. The method achieves evasion rates of at least 18% against target detectors through the generated adversarial examples. The discrete adversarial actions comprise functionality-preserving modifications to executable files including file compression, section renaming, timestamp modification, checksum alteration, signature removal, import table modification, debug information removal, section addition, and overlay appending."
Disclaimer: Curated by HT Syndication.
