ABSTRACT
In this chapter, we propose a novel testing methodology, referred to asMERO (Multiple Excitation of Rare Occurence) protection against hardware Trojans. MERO comprises of a statistical test generation approach for Trojan detection and a coverage determination approach for quantifying the level of trust. The main objective of the proposed methodology is
to derive a set of test patterns that is compact (minimizing test time and cost), while maximizing the Trojan detection coverage. The basic concept is to detect rare or low probability conditions at the internal nodes, and then derive an optimal set of vectors than can trigger each of the selected low probability nodes individually to their rare logic values multiple times (e.g. at least N times, where N is a given parameter). As analyzed in Section 14.2.1, this increases the probability of detection of an arbitrary Trojan instance. By increasing the toggling of nodes that are random-pattern resistant, it improves the probability of activating an unknown Trojan compared to purely random patterns. The proposed methodology is conceptually similar to the N-detect test [31, 302] used in stuck-at ATPG (automatic test pattern generation), where a test set is generated to detect each single stuck-at fault in a circuit by at least N different patterns to improve test quality and defect coverage [302]. In this chapter, we focus on digital Trojans [408], which can be inserted into a design either in a design house (e.g. by untrusted CAD tool or IP) or in a foundry. We do not consider the Trojans where the triggering mechanism and/or effect are analog in nature (e.g. thermal fluctuations).
