Face spoof detection techniques: IDA and PCA

Abstract

Different biometrics are used to secure the system from different kind of attacks by unauthorized entities. But among different biometrics, face biometric is very much sensitive to spoofing attacks. One of them is Face spoofing attack. It is basically an attack where a person tries to pretend to be a valid user by falsifying data and get illegitimate access. Here photo or video of an authorized person is used to gain access in to the system. The algorithms which we have proposed are very much efficient for face spoof detection, which is based on Image Distortion Analysis(IDA) and Principle Component Analysis (PCA). In IDA we will analyze features such as Specular Reflection, Blurriness, Chromatic Moment and Color Diversity. Now these four features are concatenate together. In PCA, PCA feature vector is found by calculating the eigenvectors and eigenvalues of the data covariance matrix. This extracted feature vector for both IDA and PCA is fed into multiple SVM classifiers. Each algorithm is trained on a different group of spoof training samples and results to check whether it is a spoof face or genuine face. Then we will compare and analyse the performance for both IDA and PCA.