A New Image Enhancement-Based Framework for Spoofing Detection in Ear-Based Biometric Authentication Systems

Abstract

Biometric identification systems using ears have been gaining attention in recent decades because of their advantages compared to other physiological characteristics. Consequently, attacks known as spoofing or fraud against this type of biometrics are inevitable. Therefore, it is necessary to create techniques to prevent this type of attack since, in the literature, this subject is still not sufficiently addressed. This paper proposes a new framework with data augmentation, multifiltering, feature extraction, and fusion steps for spoofing detection in biometric identification systems based on ear recognition. The proposed method aims to increase the ability of classifiers to differentiate images of real ears from fake ears. The material was analyzed considering two of the most well-known public ear databases, UBEAR and AMI, containing images of real ears, and photographs captured from the images in these databases were considered fakes. The results obtained by the proposed material were competitive or superior compared to other methods in the literature that make up the state-of-the-art in this topic.