Cyber-physical systems rely on sensors, communication, and computing, all powered by integrated circuits (ICs). ICs are largely susceptible to various hardware attacks with malicious intents. One of the stealthiest threats is the insertion of a hardware trojan into the IC, causing the circuit to malfunction or leak sensitive information. Due to supply chain vulnerabilities, ICs face risks of trojan insertion during various design and fabrication stages. These trojans typically remain inactive until triggered. Once triggered, trojans can severely compromise system safety and security. This paper presents a non-invasive method for hardware trojan detection based on side-channel power analysis. We utilize the dynamic power measurements for twelve hardware trojans from IEEE DataPort. Our approach applies to signal processing techniques to extract crucial time-domain and frequency-domain features from the power traces, which are then used for trojan detection leveraging Artificial Intelligence (AI) models. Comparison with a baseline detection approach indicates that our approach achieves higher detection accuracy than the baseline models used on the same side-channel power dataset.

翻译：信息物理系统依赖于传感器、通信与计算模块，这些模块均由集成电路供电。集成电路极易受到各类恶意硬件攻击。其中最隐蔽的威胁之一是在集成电路中植入硬件木马，导致电路功能异常或敏感信息泄露。由于供应链的脆弱性，集成电路在设计与制造的各个阶段均面临木马植入的风险。此类木马通常保持休眠状态直至被触发。一旦激活，木马可能严重危害系统的安全性与可靠性。本文提出一种基于侧信道功耗分析的非侵入式硬件木马检测方法。我们采用IEEE DataPort中十二种硬件木马的动态功耗测量数据。该方法应用信号处理技术从功耗轨迹中提取关键的时域与频域特征，随后借助人工智能模型实现木马检测。与基线检测方法的对比表明，在相同侧信道功耗数据集上，本方法获得了比基线模型更高的检测准确率。