Abstract:
Advanced system-on-chip (SoC) designs use large number of intellectual
property (IP) blocks being acquired from third-party vendors. However, a
single vulnerable IP can compromise the trustworthiness of an entire SoC,
e.g., by causing execution failures, degrading performance or leaking some
inside information. One of the root causes for such vulnerabilities are malicious alterations known as hardware trojans. In this paper, we provide a
formal methods based framework to monitor and detect the hardware trojans
at runtime. Unlike existing solutions to the untrusted IP problem, which rely
on verification of an IP before their integration into an SoC, our approach is
based on the idea of runtime resilience. In order to demonstrate its effectiveness, we apply our proposed framework to a couple of standard benchmarks
(i.e., I2C Protocol and RS232). Our results show that runtime monitors are
able to detect underlying trojans with a minimalistic overhead. Our proposed
framework is inline with the recommendations of IEC 61508 and ISO 26262 –
two main international standards for industrial automation and automotive
industry, respectively.
