Enhancing Hardware Trojan Security through Reference-Free Clustering using Representatives

To ensure the integrity of the semiconductor IC supply chain, it is of utmost importance to detect Hardware Trojans. However, existing approaches, which rely on side-channel analysis often require the utilization of golden chips for verification. To tackle this challenge, a novel methodology has been proposed in this research paper, which integrates unsupervised machine learning with side-channel analysis to detect Hardware Trojans in the semiconductor IC supply chain. The proposed method leverages unique features extracted from on-chip ring-oscillator networks to identify irregularities through unsupervised clustering. The proposed method assessment conducted on FPGA chips containing inserted Trojans displayed remarkable accuracy, surpassing alternative methods with a 99% accuracy rate. The combination of Centroid-based and Hierarchal-based clustering model exhibited superior performance with a slight edge in false positive rate and f1 score. This research contributes to bolstering trust in semiconductor IC supply chains by offering a fresh perspective on Hardware Trojan detection.