Detection of Fake Samples in Multimodal Biometric Systems

The increased threat to traditional security systems using PIN, passwords and smart cards has resulted in biometric security systems replacing the conventional methods. Susceptibility of biometric systems to spoofing attacks necessitates advanced anti-spoofing techniques. In this paper, we propose a simple method for fake detection that can be applied to different biometric samples like face, iris and fingerprint. The method proposed is software-based and non-intrusive. The proposed algorithm identifies the subspace which can be used to distinguish the real from the fake biometric samples. The projections of the training samples onto this subspace is used to train a simple Support Vector Machine (SVM) classifier. The algorithm is not specific to any biometric sample or any particular spoofing method. Another added advantage of the proposed method is its low complexity as it does not require extensive feature extraction techniques. The proposed algorithm has been validated on publicly available databases of face, iris and fingerprint.