An investigation of biometric liveness detection using various techniques

Biometric systems are growingly used for authentication purposes in various security applications. Face, Fingerprint and Iris are the biometric modalities commonly used in the authentication systems. Simplicity of use, reliability and uniqueness are important characteristics of biometric based authentications. This kind of techniques can solve the typical problem of systems based on use of password which can be forgotten or stolen. Biometric systems plays a significant role in personal, national, and global security. Despite the increase in usage of biometric based systems, these are quiet susceptible to sophisticated spoofing attacks. Various spoofing attacks are present to defeat such biometric authentication systems. To improve the level of security, it is necessary to augment the reliable liveness detection tools as software modules along with the existing authentication systems. This paper targets the countermeasures for the biometric spoofing attacks and also suggests technical measures for implementing the biometric liveness detection systems. The research in this field is very active, with local descriptors. In order to overcome the shortcomings of already existing liveness detection tools, in this work we propose two different feature extraction techniques for software-based liveness detection: Convolutional Networks and Local Binary Patterns. Both techniques were used in conjunction with a Support Vector Machine (SVM) classifier. Dataset Augmentation was used to increase classifier's performance and a variety of preprocessing operations were tested, such as frequency filtering, contrast equalization, and region of interest filtering.