MUMBAI, India, July 13 -- Intellectual Property India has published a patent application (202641083839 A) filed by Vellore Institute Of Technology, Chennai on July 08, 2026, for System For Detecting And Mitigating Adversarial Attacks And Method Thereof.
Inventors include Vishvadharman Saminathan; Yuvaraj Kurup; Fawaz Ahamed; Samarth Maligouder; Jai Vinita L; and Helen Vijitha P.
The application for the patent was published on July 10, 2026, under issue no. 28/2026.
Abstract: The present disclosure discloses a system (100) and a method (200) for detecting and mitigating adversarial attacks in a federated learning environment. The system (100) includes a plurality of distributed client devices (102) and a processing system (104) having a processor (106-1) and a memory (106-2). The processor (106-1) receives model updates from the plurality of distributed client devices (102) and converts each received model update into a structured numerical update vector using an update extraction unit (104-1). Pairwise similarity values are computed using a similarity analysis unit (104-2) to generate an inter-client similarity matrix. A swarm detection unit (104-3) identifies clusters of client devices representing coordinated behaviour. A reputation weighting unit (104-4) assigns adaptive aggregation weights. A global model (104-5) is generated using weighted aggregation, and a drift monitoring unit (104-6) detects model poisoning based on deviation between successive global models.
Disclaimer: Curated by HT Syndication.
