Enhancing Biometric Liveness Detection Using Trait Randomization Technique

Biometric Authentication Systems (BAS) have several security benefits over traditional password and token authentication including an inherent difficulty to copy, clone and share or distribute authentication credentials (biometric traits). Spoofing or presentation attack remains a major weakness of biometric systems and tackling it at the trait level is still challenging with several different approaches and methods applied in existing systems. In this paper, we focus on the well-known approach of Suspicious Presentation Detection (SPD) and present the Multi-Modal Random Trait Biometric Liveness Detection System (MMRTBLDS) that further mitigates spoofing or presentations attacks using randomization and combination of several different SPD detection techniques across three different modalities during trait capture. We discuss the detection of life using five distinct properties each from finger, face and eye modalities and present results from a simulation that highlights the improved security based on an impostor's inability to accurately predict the combination of trait liveness properties the system might prompt and test for during capture.