Granularity and detection capability of an adaptive embedded Hardware Trojan detection system

Abstract

With the outsourcing of the integrated circuit (IC) manufacturing, embedded systems faces threats as Hardware Trojan. This paper presents a characterization of a Hardware Trojan detection method introduced in a former work. In this work, a network of sensors is uniformly spread over the IC surface to monitor locally the inner supply voltage. By conducting an analysis by lot, the authors are able to get rid of the main problem of Hardware Trojan detection: the effect of intra-die and inter-die process variations. In this paper, an analysis of the spatial coverage of the method is made experimentally on a set of FPGA boards. From the obtained results, a modification of the used sensor is proposed as well as an adaptive distinguisher which aims at reducing the false positive rate. These two improvements are also experimentally tested and validated with the same set of FPGA boards.