Detecting Hardware Trojans in Manufactured Chips without Reference: A GMM-Based Approach

The horizontal IC design and manufacturing flow creates ample opportunities for unauthorized tampering at various stages, yet there is limited research dedicated to detecting hardware Trojans (HTs) in manufactured ICs. Practical applications are hindered by substantial challenges, including the presence of trusted ICs. This paper introduces a novel method for reference-free HT detection in manufactured ICs, leveraging Gaussian mixture models (GMMs) with power side-channel signals. Experimental results, conducted on AES-128 with various hardware Trojans, demonstrate an accuracy exceeding 95.52%. To be introduced for the first time, to the best of our knowledge, our proposed GMM-based HT detection method also provides a probabilistic framework, attributing evolving probability values to individual side-channel measurements to indicate potential HT activation. Consequently, it becomes feasible to ascertain with a high level of confidence whether a manufactured design has been tampered with an HT.