Hardware Trojan Detection Through Multimodal Image Processing and Analysis
David C Bowman, John M. Emmert · 2022 IEEE International Symposium on Smart Electronic Systems (iSES) · 2022
To reduce the cost of integrated circuit fabrication, a designer often submits layout masks to an untrusted foundry. Contingent on the sensitivity of the project, the cost-reduction benefit can become limited or eliminated due to the lack of security and control in an untrusted environment. Control of the design, components, and intellectual property is lost during the manufacturing step, giving an adversary the opportunity to compromise the integrated circuit. With low probability of detection, a malicious agent can insert extra circuitry in strategic locations and exfiltrate sensitive data or compromise functionality and reliability. The extra circuits are called hardware Trojans. Our project, the Automated Iterative Reverse Engineer, leverages image processing and computer vision techniques to compare the original designer's ideal GDS2 design layout to real-world micro-scopic images of the fabricated integrated circuit. Our previous work in Image Stitch Assembly and Multimodal Image Matching for Hardware Security provides a framework to leverage our Differential Image Transform and facilitate greater execution speed and higher quality matching between the ideal and real-world cases. As our data shows, the high-quality matching yields reliable mismatched areas to better indicate the presence of hardware Trojans.