Publication

Abstract : With the continued scaling of Very Large Scale Integration (VLSI) technology and the globalized Integrated Circuit (IC) supply chain, hardware security has become a major concern. Hardware Trojans are small, hidden modifications that can alter circuit behavior or leak data that pose a significant threat, and detecting them at the gate level remains difficult due to their stealthy nature and the growing complexity of designs. This work presents a lightweight ensemble learning approach for pre-silicon Trojan detection. Rather than using more classifiers, the proposed framework combines a small, optimized set of classifiers to achieve high accuracy with low computational cost. The method attains 99.50% accuracy on the CAS Lab dataset and 99.80% on the HT-PRED dataset, outperforming individual classifiers and previous ensemble techniques. These results demonstrate an efficient and scalable solution for gate-level hardware Trojan detection.