人工智能驱动型硬件木马植入检测
Artificial Intelligence Driven Hardware Trojan Implantation Detection

为应对集成电路中硬件木马植入带来的安全威胁本研究提出一种人工智能驱动型检测方法，旨在突破传统检测技术在覆盖率与未知威胁识别上的局限性。通过构建基于行为特征分析的检测框架结合卷积神经网络与动态时序建模，提取电路功耗、逻辑状态等多维度特征实现对硬件木马的精准识别。实验结果表明该方法在已知木马类型检测中实现高准确率，低误报率，且在对抗低触发概率木马及多节点协同攻击等复杂场景下表现出显著鲁棒性。与传统方法相比检测效率大幅度提升验证了人工智能技术在硬件安全领域的应用潜力。本研究为集成电路供应链安全防护提供了创新解决方案，并为未来智能检测技术轻量化与多模态融合研究奠定基础。
To address the security threats posed by hardware trojans in integrated circuits, this study proposes an AI-driven detection method aimed at overcoming the limitations of traditional detection techniques in coverage and unknown threat recognition. By constructing a detection framework based on behavioral feature analysis combined with convolutional neural networks and dynamic temporal modeling, the method extracts multi-dimensional features such as circuit power consumption and logic states to achieve precise identification of hardware trojans. Experimental results show that the method achieves high accuracy and low false positive rates in detecting known types of trojans, and demonstrates significant robustness in complex scenarios such as combating low-probability-of-attack trojans and multi-node coordinated attacks. Compared to traditional methods, the detection efficiency is significantly improved, validating the potential of AI technology in the field of hardware security. This study provides an innovative solution for supply chain security in integrated circuits and lays the foundation for future research on lightweight and multimodal fusion of intelligent detection technologies.