A Comprehensive Machine-Learning Framework for Detecting Hardware Trojans Through Enhanced Side-Channel Analysis Techniques

The globalization of the semiconductor supply chain and the distributed nature of chip design and manufacturing processes have exposed integrated circuit (IC) security to increasingly complex threats. One of the most critical among these is the insertion of malicious hardware modifications, known as hardware Trojans, during the design or fabrication stages—a threat that is often extremely difficult, or even impossible, to detect using conventional verification methods. In this study, a side-channel power analysis-based method is proposed to detect hardware Trojans. Using the Chip Whisperer board and its associated software environment, cryptographic algorithms were executed on a chip in both clean and infected versions, and their power consumption was precisely recorded. Experiments were carried out under fully controlled conditions with diverse random inputs and keys to generate a comprehensive and reliable dataset. The collected data were preprocessed through normalization, dimensionality reduction, and key feature selection, and then used to train an artificial intelligence model based on the Support Vector Machine (SVM) algorithm. Experimental results demonstrate that the proposed machine learning-based approach achieves a ۹۹.۷۳% detection accuracy, effectively identifying hardware Trojans under near-realistic conditions. Compared to conventional methods, the proposed approach provides significantly higher detection precision and reliability.

Keywords:

hardware Trojan , side-channel analysis , power consumption , support vector machine (SVM) , Chip Whisperer