Details
Logic locking is one of the most promising design-for-trust techniques for protecting intellectual property from reverse engineering, IP piracy, and modification throughout the electronic supply chain. Nowadays, even oracle-less attack is prominently used by untrusted entities in the industry for aiding reverse engineering, IP piracy, and obtaining the secret key of locked design. This demands the necessity to determine the extent of vulnerability possible on obfuscated circuitry. In this paper, we propose an attack model to break the security of logic locking by extracting the activation key of the locked netlist using a deep learning model. Our propose DeepAttack is an oracle-less attack that does not require a golden circuit for reference and has been evaluated on different circuits of ISCAS-85 and EPFL benchmarks. The algorithm can predict average key prediction accuracy (KPA) of 93.39% which is 21.28, 10.73, and 3.84 percent more than the recently proposed oracle-less attacks like SAIL, Snapshot, and OMLA respectively.