MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202631070889 A) filed by Dr. Prasanta Pratim Bairagi on June 08, 2026, for Fsf-Lids: Epsilon-Constraint Multi-Objective Rank-Based Feature Selection Framework For Lightweight Iot Ids.

Inventors include Dr. Prasanta Pratim Bairagi; Dr. Hiten Choudhury; Dr. Prakash Chauhan; Mr. Anindya Parashar; and Mr. Rafiqul Islam.

The application for the patent was published on June 26, 2026, under issue no. 26/2026.

Abstract: ABSTRACT FSF-LIDS: Epsilon-Constraint Multi-Objective Rank-Based Feature Selection Framework for Lightweight IoT IDS The present invention relates to an adaptive epsilon-constraint Multi-Objective rank-based hybrid feature selection framework for lightweight Internet of Things (IoT) intrusion detection. The framework comprises a data acquisition and preprocessing module configured to collect and preprocess IoT communication data, and a feature ranking and rank-based candidate list generation module configured to generate rank-based candidate feature subsets using multiple feature-ranking mechanisms and statistical feature analysis. A machine learning-based feature subset evaluation module evaluates generated candidate feature subsets based on intrusion detection accuracy, model training time, and model size parameters. An epsilon-constraint multi-objective optimization engine identifies feasible feature subsets using utopian-point estimation, nadir-point determination, interpolation-based epsilon-constraint generation, and adaptive tolerance factor evaluation. A multi-objective scoring and filtered feature subsets generation module calculates normalized cumulative scores corresponding to feasible feature subsets, while an optimal feature count and feature subset selection module identifies optimal feature subset suitable for resource-constrained IoT deployment environments. The integrated feature selection and optimization framework improves feature subsets quality, reduces computational overhead, minimizes model training complexity, optimizes memory utilization, and facilitates generation of lightweight intrusion detection models for IoT communication systems.

Disclaimer: Curated by HT Syndication.