Current Hardware Trojan (HT) detection techniques are mostly developed based on a limited set of HT benchmarks. Existing HT benchmarks circuits are generated with multiple shortcomings, i.e., i) they are heavily biased by the designers' mindset when they are created, and ii) they are created through a one-dimensional lens, mainly the signal activity of nets. To address these shortcomings, we introduce the first automated reinforcement learning (RL) HT insertion and detection framework. In the insertion phase, an RL agent explores the circuits and finds different locations that are best for keeping inserted HTs hidden. On the defense side, we introduce a multi-criteria RL-based detector that generates test vectors to discover the existence of HTs. Using the proposed framework, one can explore the HT insertion and detection design spaces to break the human mindset limitations as well as the benchmark issues, ultimately leading toward the next-generation of innovative detectors. Our HT toolset is open-source to accelerate research in this field and reduce the initial setup time for newcomers. We demonstrate the efficacy of our framework on ISCAS-85 benchmarks and provide the attack and detection success rates and define a methodology for comparing our techniques.

翻译：当前硬件木马（HT）检测技术大多基于有限的HT基准测试集开发。现有HT基准电路在生成过程中存在多重缺陷，具体表现为：i）其设计受创建者思维定式的强烈影响，ii）仅通过以信号线活动为主的单维度视角进行构建。为解决上述问题，我们首次提出基于自动强化学习（RL）的HT插入与检测框架。在插入阶段，RL代理通过探索电路，定位最适合隐藏所插HT的物理节点。在防御层面，我们引入基于多准则RL的检测器，通过生成测试向量以发现HT的存在。借助该框架，研究者可探索HT插入与检测的设计空间，突破人类思维局限及基准测试问题，最终推动下一代创新型检测器的发展。为加速该领域研究并降低新手入门成本，我们的HT工具集完全开源。通过ISCAS-85基准测试验证了框架的有效性，给出攻击与检测成功率，并定义了技术对比方法。