Abstract
Timely detection of Hardware Trojans (HTs) has become a major challenge for secure integrated circuits. We present a run-time methodology for HT detection that employs a multi-parameter statistical traffic modeling of the communication channel in a given System-on-Chip (SoC), named as SIMCom. The main idea is to model the communication using multiple side-channel information like the Hurst exponent, the standard deviation of the injection distribution, and the hop distribution jointly to accurately identify HT-based online anomalies. At design time, our methodology employs a "property specification language" to define and embed assertions in the RTL, specifying the correct communication behavior of a given SoC. At run-time, it monitors the anomalies in the communication behavior by checking the execution patterns against these assertions. The experimental results show that with the combined analysis of multiple statistical parameters, SIMCom is able to detect all the benchmark Trojans (available on trust-hub) with less than 1% area and power overhead.
Introduction: The Hardware Trojan Challenge
Modern System-on-Chip (SoC) designs increasingly rely on third-party Intellectual Property (3PIP) blocks to meet aggressive time-to-market demands. This globalized supply chain introduces significant security vulnerabilities, most notably Hardware Trojans (HTs)—malicious modifications that can leak sensitive data, degrade performance, or disable system functions. Traditional detection methods often struggle with "black-box" 3PIPs where a "golden" reference is unavailable, or with stealthy Trojans that remain dormant during initial testing and only activate during runtime.
The SIMCom Framework
We propose SIMCom (Statistical Sniffing of Inter-Module Communications), a runtime detection methodology that monitors the statistical "fingerprint" of internal SoC traffic. Unlike physical side-channel analysis (e.g., power or thermal), SIMCom analyzes the behavior of data packets on the internal communication bus.
The framework utilizes three primary statistical parameters to model "normal" behavior:
- Hurst Exponent ($H$): Captures the "burstiness" and long-range dependence of traffic. It is highly sensitive to changes in time-series data and requires few observations for accuracy.
- Spatial Hop Distribution ($P$): Tracks packet travel distance across the SoC topology.
- Spatial Injection ($\sigma$): Measures the standard deviation of packet injection rates.
These parameters are modeled at design time and translated into Property Specification Language (PSL) assertions embedded into the chip's RTL. At runtime, dedicated hardware monitors compute these statistics in real-time and flag any deviations from the assertions.
Experimental Evaluation
SIMCom was evaluated across three distinct SoC configurations:
- SoC1: Four single-core MC8051 microcontrollers and UART modules.
- SoC2: MC8051 cores combined with AES, Ethernet, and RSA modules.
- SoC3: Four high-end LEON3 cores connected with AES, Ethernet, and memory controllers.
To simulate attacks, we utilized standard Trust-Hub hardware Trojan benchmarks (e.g., MC8051-T200 through T800). These benchmarks represent a wide variety of payloads, including information leakage and denial-of-service (DoS) attacks.
Key Results and Insights
Experimental results demonstrate that SIMCom achieves 100% detection of the benchmark Trojans with approximately 99% accuracy. A key insight of this work is that even stealthy Trojans inevitably alter the statistical distribution of traffic when they activate, as shown in the packet distribution shifts in Figure 2.
Notably, the hardware overhead is minimal, with less than 1% area and power impact. The Hurst exponent's rapid convergence allows for fast detection, making SIMCom suitable for resource-constrained IoT devices where traditional side-channel analysis might be too slow or computationally expensive.
Conclusion and Future Work
SIMCom provides a robust, low-overhead solution for runtime HT detection in 3PIP-heavy SoCs. While it requires at least one trusted IP (such as a central CPU) to act as a "sniffer," its protocol-agnostic nature and high accuracy make it a promising defense against hardware-level supply chain attacks. Future research will explore regional monitor topologies to optimize detection coverage and the integration of automated recovery mechanisms to re-route traffic upon intrusion detection.
Summary updated