核心概念
The author presents DECOR, a method to enhance logic locking schemes against machine learning attacks by decorrelating the circuit structure from the correct key, significantly reducing key prediction accuracy.
摘要
DECOR introduces a randomized algorithm to alter UDC cofactors in locked circuits, creating one-to-many and many-to-one mappings for keys. Experimental results show a substantial decrease in key prediction accuracy with negligible advantage over random guessing. The method is applicable to various logic locking schemes and exhibits acceptable area overhead.
統計資料
Numerical results show that the proposed method can efficiently degrade the accuracy of state-of-the-art ML-based attacks down to around 50%.
The flow of DECOR involves only behavioral changes to the locked circuit and is applicable to enhance any LL technique.