Circumventing Machine Learning-Based Attacks to Logic Locking
TimeTuesday, July 12th6pm - 7pm PDT
LocationLevel 2 Lobby
DescriptionLogic Locking (LL) has gained attention as a promising IC protection measure. However, recent attacks, facilitated by machine learning (ML), have shown the potential to predict the correct key in multiple LL schemes by exploiting the correlation of possible key values with the circuit structure. This paper presents a generic construction based on a randomized algorithm that significantly decreases the correlation between locked circuit netlist and correct key values in any LL scheme. Numerical results show that the proposed method can degrade the accuracy of the state-of-the-art ML-based attack to ~50%, resulting in negligible advantage versus random guessing.