A Node Reduction Technique for Trojan Detection and Diagnosis in IoT Hardware Devices
Sree Ranjani Rajendran, Nirmala Devi, M Jayakumar · Internet of Things · 2022
Secured hardware is necessary to upgrade the performance, reliability and efficiency of any Internet of Things (IoT) system. Hardware security directly ensures the integrated circuit (IC) security, and modern electronic hardware is primarily composed of an interconnection of ICs. ICs are maliciously modified in untrusted fabrication units with Hardware Trojan (HT). These HTs are stealthy in nature and most of the detection schemes often require a golden chip for comparing the side-channel measurements made on the test IC. The existing Trojan detection self-referencing schemes have high computation complexity for bigger circuit and false-positive and false-negative detections are possible due to process variation (PV). The main objective of the proposed scheme is to obviate the need for a golden design by adopting the self-referencing scheme, which would also minimize the false-positive cases that occur due to PVs. An attempt has been made to minimize the number of power measurements by identifying specific nodes of interest by computing transition probability (TP) followed by testability measure. Then the proposed algorithm will check the consistency of the measured dynamic power at different timeframes for a set of input patterns. The inconsistency in the power measurements will detect the presence of HT at the specific node of interest; thus, identification of an important node will easily detect and diagnose HT. The main advantage of this work is a reduced time complexity on an average of 54.6%. Node reduction based on TP and SCOAP measurements increases detection accuracy with minimal power measurements. An automated design flow for the detection and diagnosis of HTs on ISCAS'85 benchmark circuits has been experimentally verified using Digilent Anvyl Spartan-6 field-programmable gate array (FPGA) trainer board. The significance of the work has reflected on the improved detection rate value with reduced detection complexity by 53.79%.