Biometric system vulnerabilities: A typology of metadata

Abstract

This study presents a root cause analysis of biometric vulnerabilities and provides a comprehensive typology of metadata in biometric adaptation. Although they are more reliable and secure than traditional authentication methods, biometric techniques are subject to vulnerabilities that pose challenges. Faced with the proliferation of cases of identity theft and fraud, biometrics is increasingly used to protect assets and people in several areas such as commercial, forensic and government applications. As a first step, a metadata analysis was performed. A focus has then been placed on their role in the fight against biometric vulnerabilities. Thus, the vulnerabilities studied have been classified into two main categories: intrinsic limits and adverse attacks. Finally, one of the scenarios considered was implemented, particularly the case of the combination of skin color with facial recognition. The implementation resulted in encouraging results with an Area Under the Curve (AUC) of 0.826 for the face system and 0.908 for the multimodal system.