MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202641077869 A) filed by Srm Institute Of Science And Technology, Ramapuram Campus; and Easwari Engineering College on June 24, 2026, for Netwatch: Interpreting Network Anomalies Through Explainable Unsupervised Behavioral Modeling.
Inventors include Manav Raitani; Ishan Tripathi; Jakka Nomith Ujjwal; and Dr. R. Bhuvaneswari.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: Existing network anomaly detection systems suffer from static baselines that degrade under traffic regime shifts, black-box scoring that provides no operational context, and flow-level granularity that overwhelms analysts with undifferentiated alerts. Extending the explainable anomaly detection framework of Shafir, Giryes, and Wool [24], this paper presents a self-calibrating behavioral detection system that addresses each limitation through four novel mechanisms: time-segmented baseline modeling across four diurnal periods, periodic soft baseline refresh that retrains the detection model every fifteen minutes without interrupting detection, confidence- weighted anomaly thresholding that suppresses false positives during baseline immaturity, and counterfactual explanation generation that produces per-anomaly threshold statements specifying the exact behavioral change that would have prevented each alert. The proposed framework requires no labeled training data, predefined attack signatures, or manual recalibration, enabling autonomous and interpretable anomaly detection on live network traffic without prior knowledge of attack characteristics.
Disclaimer: Curated by HT Syndication.
