MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202641076514 A) filed by Cvr College Of Engineering on June 20, 2026, for Enhanced Anomaly Detection In Cyber-Physical Systems Via Spectrogram Optimization And Hybrid Learning Models.
Inventor includes B. Sharmila.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: Integrated Cyber-Physical Systems in critical infrastructure improve automation but face cybersecurity threats, necessitating robust attack detection. Current detection models, while effective, struggle with issues like overfitting, scalability, and time consumption. This study introduces an anomaly-based detection system using optimized spectrogram features and a hybrid machine learning approach for CPS security. It utilizes Synthetic Minority Oversampling Technique for data augmentation, Binary Lévy Flight Osprey Optimization for feature selection, and Short-Time Fourier Transform for spectrogram image creation. Multi-scale features are then extracted from these images using a Pyramid Dilated Deeper Encoder–Decoder model. The attack detection employs a Hybrid Secretary Optimized Radial Basis Support Vector Machine-based Extreme Gradient Boosting method, optimizing kernel parameters with the Secretary Bird algorithm. Experimentally, this model outperforms existing methods with an accuracy of 99.007% on HAI Security and 99.108% on CIC-IDS2017 datasets.
Disclaimer: Curated by HT Syndication.
