SALTY: Explainable Artificial Intelligence Guided Structural Analysis for Hardware Trojan Detection

Hardware Trojans pose serious real-world risks: Malicious modifications hidden in computer chips can enable data theft or cause system failures in critical infrastructure like healthcare and aviation systems, potentially endangering lives and causing major financial losses.

Previous detection methods had significant limitations: Existing techniques for finding these hidden threats suffered from scalability problems and generated excessive false alarms, which could undermine trust in security systems by creating alert fatigue.

AI explainability improves detection reliability: SALTY combines machine learning predictions with explainable AI (XAI) techniques to verify and refine results, achieving 98% accuracy in both correctly identifying threats and correctly clearing safe designs—a substantial improvement over previous methods.

Graph Neural Networks are well-suited for this problem: The framework uses a specialized neural network architecture designed to analyze the structural relationships within chip designs, suggesting that understanding how components connect is key to spotting suspicious modifications.

Supply chain security requires trustworthy verification: Since chips often come from third-party vendors, having reliable detection methods with transparent reasoning (via XAI) is essential for organizations to confidently verify that purchased components haven't been compromised before integration into critical systems.