Ingress of Threshold Voltage-Triggered Hardware Trojan in the Modern FPGA Fabric-Detection Methodology and Mitigation

Abstract

The ageing phenomenon of negative bias temperature instability (NBTI) continues to challenge the dynamic thermal management of modern FPGAs. Increased transistor density leads to thermal accumulation and propagates higher and non-uniform temperature variations across the FPGA. This aggravates the impact of NBTI on key PMOS transistor parameters such as threshold voltage and drain current. Where it ages the transistors, with a successive reduction in FPGA lifetime and reliability, it also challenges its security. The ingress of threshold voltage-triggered hardware Trojan, a stealthy and malicious electronic circuit, in the modern FPGA, is one such potential threat that could exploit NBTI and severely affect its performance. The development of an effective and efficient countermeasure against it is, therefore, highly critical. Accordingly, we present a comprehensive FPGA security scheme, comprising novel elements of hardware Trojan infection, detection, and mitigation, to protect FPGA applications against the hardware Trojan. Built around the threat model of a naval warship's integrated self-protection system (ISPS), we propose a threshold voltage-triggered hardware Trojan that operates in a threshold voltage region of 0.45V to 0.998V, consuming ultra-low power (10.5nW), and remaining stealthy with an area overhead as low as 1.5% for a 28 nm technology node. The hardware Trojan detection sub-scheme provides a unique lightweight threshold voltage-aware sensor with a detection sensitivity of 0.251mV/nA. With fixed and dynamic ring oscillator-based sensor segments, the precise measurement of frequency and delay variations in response to shifts in the threshold voltage of a PMOS transistor is also proposed. Finally, the FPGA security scheme is reinforced with an online transistor dynamic scaling (OTDS) to mitigate the impact of hardware Trojan through run-time tolerant circuitry capable of identifying critical gates with worst-case drain current degradation.