Face Spoofing Detection using Color Distortion Features and Principal Component Analysis

Face anti-spoofing is an important topic of face recognition system to protect against security breach. Previous approach for face spoofing detection based on distortion in images have achieved promising results. However, their generalization ability has not been sufficiently addressed. In this work, we propose a face spoofing detection based on color distortion analysis, which captures the chromatic aberration from a face image. Color distortion analysis extracts color moment and ranked histogram features, which generate 116 feature vector. The feature vector then forwarded to Principal Component Analysis (PCA) to perform dimensionality reduction. For classifying a live or spoof face image, a Naïve Bayes classifier performed on the principal components obtained from PCA. From experiment, the proposed method achieves competitive performance compared to previous approach, with the highest TPR (True Positive Rate) is 97.4%.