TRAP-GATE: A Probabilistic Approach to Enhance Hardware Trojan Detection and its Game Theoretic Analysis
TRAP-GATE: A Probabilistic Approach to Enhance Hardware Trojan Detection and its Game Theoretic Analysis
复制标题
TRAP-GATE：一种增强硬件木马检测的概率方法及其博弈论分析
DOI：
10.1007/s10836-020-05907-z
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Chandra Babu
中科院分区：
文献类型：
--
作者：
Sivappriya Manivannan;L. Kuppusamy;N. Sarat;Chandra Babu
从恶意软件攻击中拯救硬件是当今的一大挑战。此外，使用低成本的技术检测恶意入侵的存在是非常具有挑战性的，特别是当它被认为是硬件木马集成到很少兴奋的节点。虽然逻辑测试被认为是检查电路功能正确性的准确方法，但对于需要大量测试模式来激活硬件木马的电路，该方法变得低效。本文提出了一种概率的方法来生成一组测试模式，通过观察应用的测试模式的不正确的反应，明确激活木马。我们在ISCAS'85基准电路上的实验结果表明，我们的方法生成的测试模式包含所有的网络，有很高的机会木马插入。我们强调，虽然我们的结果是特定的电路，所提出的方法是通用的，因此它可以应用于任何数字电路。事实上，我们证明，对于C880电路，我们的方法只需要217输入进行测试，而天真的方法需要260测试模式。除了实验结果，我们使用一个博弈论的框架，以显示我们的方法在生成木马激活测试模式相比，天真和ATPG测试过程的有效性。
Rescuing Hardware from malware attacks is a great challenge today. Moreover detecting the presence of malicious intrusion using low-cost techniques is very challenging especially when it is believed that hardware Trojans are integrated into the rarely excited nodes. Though logical testing is admitted to be the accurate way to check the functional correctness of the circuit, the method becomes inefficient for circuits requiring humongous number of test patterns to activate the hardware trojan. This paper proposes a probabilistic approach to generate a set of test patterns that activate the Trojan explicitly by observing the incorrect responses to the applied test patterns. Our experimental result in ISCAS’85 benchmark circuits show that the test patterns generated by our approach enclose all nets where there is a high chance for trojan insertion. We stress that though our results are circuit specific, the proposed approach is generic and hence it can be applied to any digital circuit. In fact, we demonstrate that for the C880 circuit, our approach requires only 217inputs to be tested whereas the naive approach needs 260test patterns. In addition to the experimental results, we use a game-theoretic framework to show the effectiveness of our approach in generating trojan activating test patterns compared to naive and ATPG testing process.