MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202621078975 A) filed by Indian Institute Of Technology Gandhinagar on June 26, 2026, for A Method For Evaluating Stability And Robustness Of Feature Representations In Cybersecurity Analytics.
Inventors include Sharma, Yash; and Kulkarni, Sameer Gundrao.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: ABSTRACT A METHOD FOR EVALUATING STABILITY AND ROBUSTNESS OF FEATURE REPRESENTATIONS IN CYBERSECURITY ANALYTICS The present disclosure relates to a method for evaluating the stability and robustness of feature representations in cybersecurity analytics. A classical processor (110) retrieves and preprocesses a cybersecurity dataset to extract a reduced feature vector, generating a baseline classical representation. This vector is then transmitted to a quantum processor (116) to construct two distinct quantum embeddings: an angle-encoded quantum state and a Hamiltonian-inspired quantum state utilizing alternating cost and mixer operators. The classical processor (110) injects controlled perturbation models such as additive Gaussian, bounded uniform, or L2-bounded noise into test samples exclusively during the inference phase, avoiding the need to retrain the classifier. A comparator circuit (120) evaluates the robustness of the classical, angle-encoded, and Hamiltonian-inspired representations by comparing their predictive performance and geometric stability under both clean and noisy condition.
Disclaimer: Curated by HT Syndication.
