Publication

Abstract : VLSI technology has advanced significantly, with ICs now containing billions of transistors. Their security and reliability are of utmost importance because Hardware Trojans (HTs), malicious modifications to Integrated Circuits (ICs), pose serious threats to their integrity and functionality. This work classifies the circuits into trojan-free and trojan affected using an Adaptive Genetic Algorithm (AGA) to optimize feature selection. The application of AGA enables choosing the most discriminative and relevant features out of the available features, eliminating redundancy and noise within the dataset, thus enhancing the learning process of the models. In addition, genetic algorithms are also implemented in hyperparameter tuning of supervised learning methods, which further enhances their performance by determining optimal settings for classification problems. This double use of genetic algorithms both for hyperparameter tuning and feature selection increases not only the performance of the detection system but also its accuracy. This approach gives an improvement in accuracy up to 98% and improvement in TNR when compared with normal supervised learning techniques.