Cost-efficient Chip Identification Method using Scan Flip-flop based Physically Unclonable Function

Abstract

Scan flip-flop based physically unclonable function (SCAN-PUF) has been proposed to protect integrated circuits (ICs) from security threats such as unauthorized access and IC cloning. In this paper, we propose an efficient SCAN-PUF technique that improves the uniqueness of responses with low cost overhead. The proposed SCAN-PUF first determines an optimal number of power-up state observations and then selects scan flip-flops as the PUF elements (P-ELEMENTs) through a given number of observations. A Bayesian model is adopted to evaluate the reliability of the P-ELEMENTs, and a grouped P-ELEMENT selection method is introduced to obtain more P-ELEMENTs than a predetermined threshold. To evaluate the proposed SCAN-PUF, we observed the power-up states of scan flip-flops from 15 chips fabricated using the 65-nm CMOS technology. The optimal number of observations is determined according to the reliability of the P-ELEMENTs, and the reliability, randomness, and uniqueness of the responses are then analyzed.