Cluster-BPI: Efficient Fine-Grain Blind Power Identification for Defending against Hardware Thermal Trojans in Multicore SoCs

TL;DR

This work tackles the challenge of fine-grained power estimation and thermal management in multicore SoCs by enhancing Blind Power Identification (BPI) with Improved Clustering BPI (ICBPI). By using DBSCAN to initialize the thermal resistance matrix R, ICBPI improves NNMF convergence, yielding more accurate per-core power estimates and more robust detection of thermal sensor attacks. The approach achieves up to 77.56% lower error than BPI and 68.44% lower than BPISS, and it strengthens the BIC security framework by virtually eliminating detection failures across tested conditions. Overall, ICBPI advances both thermal management accuracy and hardware security for heterogeneous multicore platforms, though simultaneous attack scenarios remain a challenge.

Abstract

Modern multicore System-on-Chips (SoCs) feature hardware monitoring mechanisms that measure total power consumption. However, these aggregate measurements are often insufficient for fine-grained thermal and power management. This paper presents an enhanced Clustering Blind Power Identification (ICBPI) approach, designed to improve the sensitivity and robustness of the traditional Blind Power Identification (BPI) method. BPI estimates the power consumption of individual cores and models the thermal behavior of an SoC using only thermal sensor data and total power measurements. The proposed ICBPI approach refines BPI's initialization process, particularly improving the non-negative matrix factorization (NNMF) step, which is critical to the accuracy of BPI. ICBPI introduces density-based spatial clustering of applications with noise (DBSCAN) to better align temperature and power consumption data, thereby providing more accurate power consumption estimates. We validate the ICBPI method through two key tasks. The first task evaluates power estimation accuracy across four different multicore architectures, including a heterogeneous processor. Results show that ICBPI significantly enhances accuracy, reducing error rates by 77.56% compared to the original BPI and by 68.44% compared to the state-of-the-art BPISS method. The second task focuses on improving the detection and localization of malicious thermal sensor attacks in heterogeneous processors. The results demonstrate that ICBPI enhances the security and robustness of multicore SoCs against such attacks.

Figures (8)

• Figure 1: Interactions and data-flow of the Blind Power Identification Algorithm
• Figure 2: Illustration of the DBSCAN Clustering Algorithm geron2022machine
• Figure 3: K-distance graph to determine the value of $\epsilon$
• Figure 4: ICBPI Custom Dataset Generation and Validation Workflow.
• Figure 5: The simulation models a 4-core processor where thermal sensor '1' is made malicious by introducing an error to its reading, $t_1$.