Supervised Machine Learning Techniques for Trojan Detection with Ring Oscillator Network

Abstract

With the globalization of the semiconductor manufacturing process, electronic devices are powerless against malicious modification of hardware in the supply chain. The everincreasing threat of hardware Trojan attacks against integrated circuits has spurred a need for accurate and efficient detection methods. Ring oscillator network (RON) is used to detect the Trojan by capturing the difference in power consumption; the power consumption of a Trojan-free circuit is different from the Trojan-inserted circuit. However, the process variation and measurement noise are the major obstacles to detect hardware Trojan with high accuracy. In this paper, we quantitatively compare four supervised machine learning algorithms and classifier optimization strategies for maximizing accuracy and minimizing the false positive rate (FPR). These supervised learning techniques show an improved false positive rate compared to principal component analysis (PCA) and convex hull classification by nearly 40% while maintaining > 90% binary classification accuracy.

AI-enabled Hardware Trojan Detection for Secure and Trusted Context-Aware Embedded Systems

“…This assessment provides insights into the scalability and robustness of the models across varying sample sizes, offering valuable implications for the practical applicability of the detection methods in real-world scenarios with diverse data availability. Moreover, this approach facilitates a direct comparison of results with the supervised learning techniques introduced by [10], employing the same evaluation method and dataset. Each sample size underwent 20 trials, and metrics including average accuracy, false positive rate (FPR), false negative rate (FNR), true negative rate (TNR), and true positive rate (TPR) were calculated and recorded.…”

“…3: Configuration of the ring oscillator network employed for Trojan detection. The structure may vary based on the power network of the specific Integrated Circuit (IC) under protection, even though N Ring Oscillators (ROs) are utilized in this particular setup [10], [11] structure and patterns within unlabeled datasets. These algorithms excel in scenarios where large datasets require efficient organization and clustering, without the need for prior labeling or categorization of the data points.…”

“…Machine Learning, combined with Design-for-Trust, offers an additional Side-Channel Analysis tactic. This involves creating diverse models using signatures from integrated com- [10], achieving an impressive 94% accuracy in distinguishing hardware trojans using multiple supervised machine learning models. However, this approach relies on labeled data, necessitating Golden References, which may be costly and inefficient in an industrial application.…”

A Golden-Free Unsupervised ML-Assisted Security Approach for Detection of IC Hardware Trojans

“…specific signature. An illustrative case is demonstrated in the work by K. Worley et al in [26], where they have implemented supervised machine learning models (a voting ensemble of K-Nearest Neighbors, Support Vector Machine, and Naïve Bayes) attaining an impressive accuracy rate of 94% in distinguishing whether an IC chip is contaminated with a hardware trojan or not. However, it's important to note that this approach still relies on the necessity for labeled data, implying the requirement for Golden References (Golden Chips) which then provide the RO Frequency signatures as labeled data for IC chips unaffected by hardware trojans.…”

Using decision tree classifier to detect Trojan Horse based on memory data

“…Worley and Rahman [22] conducted a quantitative assessment comparing the effectiveness of four different supervised ML techniques in classifying integrated circuits based on their ring oscillator network frequencies. Remarkably, when utilizing an SVM classifier, this approach achieved 97.6% accuracy in binary classification, accompanied by an impressively low false positive rate (FPR) of just 7.1%.…”