학술논문

BIC: Blind Identification Countermeasure for Malicious Thermal Sensor Attacks in Mobile SoCs
Document Type
Conference
Source
2022 23rd International Symposium on Quality Electronic Design (ISQED) Quality Electronic Design (ISQED), 2022 23rd International Symposium on. :1-6 Apr, 2022
Subject
Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Photonics and Electrooptics
Temperature sensors
Temperature distribution
Runtime
Quality of service
Benchmark testing
Thermal management
Steady-state
Language
ISSN
1948-3295
Abstract
Mobile System-on-Chips (SoCs) heavily rely on dynamic thermal management (DTM) methods in order to deal with their thermal and power density issues at runtime. The efficiency of any DTM method is directly related to the temperature data coming from the thermal sensors. For the first time, in this paper, we introduce a serious security attack on thermal sensors that can alter both the performance and reliability of the chip. We propose a Blind Identification Countermeasure (BIC) that successfully defeats the attack by identifying and isolating the infected sensor. In addition, the proposed method can accurately estimate the steady state temperature of the core associated with the isolated thermal sensor so that the DTM can continue its services with no interruption. Based on our wide range of evaluations, BIC can provide an excellent accuracy of 100% in detecting attacking sensors with a maximum temperature estimation error of ≈0.18°C. Also, BIC inflects a negligible performance overhead of 0.7% when tested with Geekbench 4.3.1 benchmark suite.