Confidence-Based Rare-Net Fusion with MUX-Based Checker for Hardware Trojan Detection

Abstract

Hardware Trojans (HTs) are among the most serious threats to the security of integrated circuits (ICs). Detection of Hardware Trojans is not easy due to the extremely low probability of activation of trojan trigger nets. Further, the Trojan circuitry bears a structural similarity to the original circuit logic. This paper uses rare-net analysis to identify the internal nets that are strong candidates for Hardware Trojan insertion. Runtime monitoring provides the final line of defense through the continuous monitoring of a circuit’s behavior during operation. This work proposes a detection framework that integrates rare-net analysis with runtime monitoring. The proposed detection framework employs two complementary schemes called Composite Rarity Score (CRS) and Hierarchical Multi-Stage Rarity Framework (HMRF) to identify possible Trojan locations. The techniques assess different and complementary characteristics of circuit rarity. The framework combines their outputs through confidence-based fusion to obtain a unified set of high-confidence rare nets. ATPG creates test patterns specifically for the selected high-confidence rare-nets. A greedy selection algorithm removes redundant test patterns while preserving the activation coverage of all high-confidence rare nets. A MUX-based checker (with XOR comparison) is used for runtime monitoring to detect deviations from the expected (golden) circuit behavior. The proposed framework is evaluated on ISCAS benchmark circuits achieving up to 75% monitoring-net reduction while maintaining high relevance to Trojan locations and a reliable checker performance over the benchmark circuits of varying complexity.