Deep Learning in Biometric Authentication: Challenges, Recent Advancements, and Future Trends

Abstract

Biometric systems have significantly improved with the integration of machine learning, especially with deep learning methods. As technology advances, it presents both new security risks and chances for biometrics to be used in a variety of contexts. Due to their remarkable success, recently, deep learning-based biometric methods have seen a dramatic increase in research papers. Thus, to compile and analyze research publications in the field, this paper presents a systematic a Systematic Literature Review (SLR). The review adopts the Systematic Reviews and Meta-Analyses (PRISMA) approach for the collection, analysis, and reporting of findings from relevant research. For this study, 109 papers in all were chosen from top digital libraries using predetermined search criteria. The most popular deep learning approaches in these systems, important biometric authentication strategies, and commonly used assessment metrics are highlighted in the review. Finally, the main challenges with deep learning-based biometric authentication methods are outlined in the paper, along with possible future trends in the field.