INSIGHT: Attacking Industry-Adopted Learning Resilient Logic Locking Techniques Using Explainable Graph Neural Network

Watch a 12-minute conference presentation from USENIX Security '24 exploring INSIGHT, a novel attack method using explainable Graph Neural Networks (GNN) to compromise learning resilient logic locking techniques. Learn how researchers from NYU, NYU Abu Dhabi, and University of Delaware developed an oracle-less approach that successfully recovers secret keys from 7 previously unbroken locking techniques, including 2 industry-adopted solutions. Discover how INSIGHT achieves significantly higher key-prediction accuracy compared to existing machine learning attacks, with demonstrations across academic test suites and real-world processors like MIPS, Google IBEX, and mor1kx. Examine practical case studies showing successful attacks against commercial EDA tools and the implications for image processing applications. Gain insights into the current state of hardware intellectual property protection and the ongoing challenges in developing truly secure logic locking techniques.