Formal Verification Approach to Detect Always-On Denial of Service Trojans in Pipelined Circuits

Abstract

Always-On Denial of Service (DoS) Trojans with power drain payload can be disastrous in systems where on-chip power resources are limited. These Trojans are designed so that they have no impact on system behavior and hence, harder to detect. A formal verification method is presented to detect sequential always-on DoS Trojans in pipelined circuits and pipelined microprocessors. Since the method is proof-based, it provides a 100% accurate classification of sequential Trojan components. Another benefit of the approach is that it does not require a reference model, which is one of the requirements of many Trojan detection techniques (often a bottleneck to practical application). The efficiency and scalability of the proposed method have been evaluated on 36 benchmark circuits. The most complex of these benchmarks has as many as 135,898 gates. Detection times are very efficient with a 100% rate of detection, i.e., all Trojan sequential elements were detected and all non-trojan sequential elements were classified as such.