MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202611065004 A) filed by Noida Institute Of Engineering And Technology Niet on May 22, 2026, for Privacy-Preserving Cyber Threat Detection System Using Transformer- Based Classification With Encrypted Network Traffic Tokenization.
Inventors include Dr. Vineet Kumar; and Mohd. Abdul Mateen Siddiqui.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: A cyber threat detection and vulnerability assessment system (100) for Industrial Internet of Things network environments addresses semantic incompatibility between numerical network traffic features and transformer word-embedding representations. The system (100) comprises a data collection module (101) receiving packet capture data, a feature extraction module (102), a cipher-shift encryption module (103) transforming numerical feature values into alphabetic ciphertext sequences, a byte-level byte pair encoding tokenizer (104) decomposing encrypted sequences into subword tokens, a preprocessing pipeline (105) applying normalization and attention-masked padding, a four-layer stacked transformer encoder (106) with multi-head self-attention and Gaussian Error Linear Unit feed-forward sublayers, and a softmax classification layer (107) classifying threats across fifteen attack categories. The cipher- shift encryption module (103) and tokenizer (104) bridge the semantic gap between numerical packet features and transformer input space while preserving data privacy for Industrial Internet of Things infrastructure security.
Disclaimer: Curated by HT Syndication.
