Trojan Detection via Side-Channel Sensitivity Maximization

The patent segments the test generation process by identifying and focusing only on suspicious regions within the circuit design. Instead of generating tests for the entire large SoC design, the method divides the circuit into suspicious and non-suspicious regions, and generates tests specifically targeted at the suspicious regions. This segmentation dramatically reduces test generation time while maintaining detection effectiveness.

The patent applies local quality by making different parts of the circuit undergo different treatment. Suspicious regions are subjected to detailed side-channel analysis and targeted test generation, while non-suspicious regions are excluded or minimized in the test generation process. This localized approach concentrates computational resources where they are most needed, improving both sensitivity and efficiency.

The patent performs preliminary action by pre-identifying suspicious regions in the circuit design before test generation. Using static analysis and heuristic methods, the system预先 marks regions that are likely to contain Trojans based on design patterns, insertion points, and risk assessment. This preliminary identification allows subsequent test generation to focus only on these marked regions, dramatically reducing complexity while maintaining detection coverage.

The patent applies partial action by generating test patterns that are sufficient to detect Trojans in suspicious regions without requiring exhaustive coverage of the entire circuit. The method generates enough test patterns to achieve high detection probability in targeted regions, rather than attempting complete coverage of all circuit paths, which would be computationally infeasible for large designs.

The patent applies parameter changes by dynamically adjusting test conditions and measurement parameters to maximize the signal from potential Trojans. The system varies input patterns, measurement timing, and analysis parameters to enhance the side-channel signatures from suspicious regions while minimizing the impact of noise and process variations. This adaptive parameter adjustment significantly improves detection sensitivity.

The patent uses copying by creating multiple copies of test patterns and applying them in sequence to the suspicious regions. By repeatedly applying similar test patterns with slight variations and aggregating the results, the system amplifies the signal from potential Trojans while averaging out random noise. This repeated measurement approach enhances sensitivity without requiring more invasive detection methods.