A Two-Stage Feature Transformation-Based Fingerprint Authentication System for Privacy Protection in IoT

Abstract

The significant and rapid development of the Internet of Things (IoT) in recent years has greatly benefited people's lives. However, there are serious security and privacy concerns that need to be addressed when using the IoT for distributed authentication. In this article, we propose a secure fingerprint authentication system to protect user privacy for authentication on IoT devices. The proposed system applies a two-stage feature transformation scheme. Specifically, a weight-based fusion mechanism is designed in the first stage, while the second stage is featured by a linear convolution-based transformation with element removal from the convolution output to increase the security and protection. The proposed authentication system satisfies all the requirements of cancelable biometrics: accuracy, revocability, and diversity, unlinkability, and noninvertibility. Evaluated over six public fingerprint databases, the proposed authentication system exhibits highly competitive performance when compared with the existing cancelable fingerprint templates. Moreover, its energy efficiency on savings in memory space and low computational costs make the proposed scheme a good fit for resource-limited IoT devices.