GALU: A Genetic Algorithm Framework for Logic Unlocking

Logic locking is a circuit obfuscation technique that inserts additional key gates to the original in order prevent potential threats such as overproduction, piracy, and counterfeiting. The encrypted generates desired outputs only when correct keys are applied gates. Previous works have identified vulnerability of logic satisfiability (SAT)-based attacks. However, SAT attacks unscalable limited effectiveness on circuits with SAT-hard structures. To address above constraints, we propose GALU, first genetic algorithm-based unlocking framework parallelizable significantly faster than conventional SAT-based counterparts. GALU by formulating deobfuscation (i.e., identifying keys) combinatorial optimization problem approaches it using algorithms (GAs). We consider sequences individuals distinct populations an adaptive, diversity-guided GA consisting four main steps: fitness evaluation, population selection, crossover, mutation. In each iteration, high scores selected transformed into offspring sequences. As result evolutionary searching, highly scalable, effective, efficient. optimize runtime overhead unlocking, integrate design GALU’s algorithm, software hardware closed loop. particular, identify evaluation performance bottleneck employ emulation programmable for optimization. this end, automatically constructs customized auxiliary circuitry pipeline computation constraints checking, sorting, adaptive scalable attack provides flexibility/trade-off between usability. This achieved producing group approximate improving quality over time. perform comprehensive various benchmarks demonstrate achieves up 1089.2× speedup 4268.6× more energy-efficiency compared state-of-the-art unlocking.