Abstract
This paper reports a novel approach that uses transistor aging in an integrated circuit (IC) to detect hardware Trojans. When a transistor is aged, it results in delays along several paths of the IC. This increase in delay results in timing violations that reveal as timing errors at the output of the IC during its operation. We present experiments using aging-aware standard cell libraries to illustrate the usefulness of the technique in detecting hardware Trojans. Combining IC aging with over-clocking produces a pattern of bit errors at the IC output by the induced timing violations. We use machine learning to learn the bit error distribution at the output of a clean IC. We differentiate the divergence in the pattern of bit errors because of a Trojan in the IC from this baseline distribution. We simulate the golden IC and show robustness to IC-to-IC manufacturing variations. The approach is effective and can detect a Trojan even if we place it far off the critical paths. Results on benchmarks from the Trust-hub show a detection accuracy of $\geq$99%.
Introduction: The Invisible Threat in Silicon
The global semiconductor supply chain has become increasingly fragmented, with design houses often outsourcing manufacturing to third-party offshore foundries. This decentralization introduces significant security risks, most notably the insertion of "Hardware Trojans"—malicious modifications to an integrated circuit (IC). These Trojans can be designed to leak sensitive information, disable functionality, or alter the chip's behavior under specific, rare conditions.
As shown in Figure 1, Trojans can be inserted at various stages of the design and fabrication process, from RTL design to the physical layout level. Because these modifications are often dormant and occupy a tiny fraction of the total circuit area, traditional structural and functional tests frequently fail to detect them.
Methodology: Amplifying Malicious Signals
The proposed detection method leverages two physical phenomena to expose hidden Trojans: Controlled Aging and Overclocking.
1. Controlled Aging and Overclocking
Transistors naturally slow down over time due to Bias Temperature Instability (BTI). By simulating this aging process using aging-aware standard cell libraries, the researchers can intentionally induce path delays. When combined with overclocking—running the IC faster than its rated frequency—these delays result in timing violations that manifest as bit errors at the output.
2. Machine Learning Pipeline
The core of the detection system is a machine learning pipeline designed to distinguish between the error patterns of a "clean" IC and one containing a Trojan.
The pipeline, illustrated in Figure 2, consists of several key stages:
- Feature Extraction: The system monitors bit flips (0→1 and 1→0) and their locations at the IC output across various aging states and clock periods.
- Feature Compactification: A four-layer Autoencoder with ReLU activations reduces the high-dimensional feature tensor into a compact vector.
- Anomaly Detection: A One-Class Support Vector Machine (SVM) is trained on features from a "golden" (clean) baseline. The SVM acts as a classifier, identifying any significant divergence in the error pattern as a potential Trojan.
Experimental Results
The approach was validated using benchmarks from the Trust-hub repository, including RSA and AES circuits. Key findings include:
- High Sensitivity: The system detected Trojans occupying as little as 0.22% of the circuit area.
- Robustness: Even when Trojans were placed thousands of paths away from the critical path (e.g., in the AES-T1000 benchmark), the method maintained high accuracy.
- Accuracy: Across all tested benchmarks, the detection accuracy remained $\geq$99%.
- Manufacturing Variation: The model proved robust against IC-to-IC manufacturing variations, successfully distinguishing malicious delays from natural process noise.
Conclusion
By using circuit aging as a natural trigger and overclocking as a magnifying glass, this research provides a powerful tool for securing the hardware supply chain. The ability to detect dormant Trojans without needing to trigger their specific logic represents a significant advancement in hardware security. Future work will focus on validating these results on physical silicon and exploring fast voltage switching as a practical alternative to simulated aging.
Paper Citation: V. R. Surabhi et al., "Hardware Trojan Detection Using Controlled Circuit Aging," arXiv:2004.02997v3 [cs.AR], 2020.
Summary updated