Deep Graph Learning for Circuit Deobfuscation

Circuit obfuscation is a recently proposed defense mechanism to protect the intellectual property (IP) of digital integrated circuits (ICs) from reverse engineering. There have been effective schemes, such as satisfiability (SAT)-checking based attacks that can potentially decrypt obfuscated circuits, which called deobfuscation. Deobfuscation runtime could be days or years, depending on layouts ICs. Hence, accurately pre-estimating deobfuscation within reasonable amount time crucial for IC designers optimize their defense. However, it challenging due (1) complexity graph-structured circuit; (2) varying-size topology circuits; (3) requirement efficiency method. This study proposes framework predicts graph deep learning techniques address challenges mentioned above. A conjunctive normal form (CNF) bipartite utilized characterize this SAT problem by analyzing attack Multi-order information matrix designed identify essential features and reduce computational cost. To overcome difficulty in capturing dynamic size CNF graph, an energy-based kernel aggregate into identical vector space. Then, we framework, Deep Survival Analysis with Graph (DSAG), integrates layers inspired censored regression survival analysis. Integrating uncensored data data, model improves standard significantly. DSAG end-to-end automatically extract determinant runtime. Extensive experiments benchmarks demonstrate its effectiveness efficiency.