MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202641061225 A) filed by Dharsana Dharani VG; and Harini M, Chennai, Tamil Nadu, on May 14, for 'transformer-based pattern recognition for obfuscation analysis and deobfuscation of web attacks using adversarial modsecurity testing and feature engineering.'
Inventor(s) include E. Uma; Rohith R; Dharsana Dharani VG; and Harini M.
The application for the patent was published on May 29, under issue no. 22/2026.
According to the abstract released by the Intellectual Property India: "The increasing use of obfuscated web attack payloads poses a major challenge to traditional Web Application Firewalls (WAFs). Attackers often use encoding techniques such as URL encoding, Base64, and Unicode to hide malicious payloads like SQL Injection (SQLi), Cross-Site Scripting (XSS), and Command Injection (CmdI). This system introduces a Transformer-based framework that first deobfuscates encoded payloads using the T5-small model and then classifies them using the DistilBERT model. The framework categorizes payloads into SQLi, XSS, CmdI, or benign classes with improved accuracy. Feature engineering techniques including entropy scoring, character frequency analysis, symbol count, payload length, and encoding pattern analysis further enhance detection reliability. A local ModSecurity WAF testbed is used to simulate real-world security environments and validate whether payloads are allowed or blocked. The dataset contains both benign and malicious payloads manually encoded using multiple techniques. This approach bridges the gap between academic research and practical web security implementation."
Disclaimer: Curated by HT Syndication.
