Abstract:
Biometric systems verify humans using their unique physiological or behavioural patterns to offer more
secure authentication over passwords and tokens. Despite their benefits, Biometric Authentication
Systems remain vulnerable to spoofing, wherein an impostor presents a forged biometric trait and
bypasses security checks. Impacts of successful spoofing can be potentially fatal such as in healthcare
and crime investigation systems where insecure authentication can result in patient misdiagnosis and
criminal misidentification, respectively. Existing anti-spoofing techniques are mostly uni-modal and
predictable, and therefore incapable of coping with the sophistication of modern-day biometric
cyberattacks. This paper presents the Multi-Modal Random Trait Biometric Liveness Detection System
(MMRTBLDS) framework which employs a complex trait randomization algorithm to mitigate
predictability. Fifteen liveness attributes derived from finger, face and iris traits are used to simulate
various authentication scenarios, resulting in 99.2% efficiency over uni-modal biometric systems. The
paper also proposes areas of useful application of the framework based on its capacity to neutralize an
impostor’s ability to accurately predict biometric trait combinations at the sensor verification stage.
