CHIF: Convoluted Histogram Image Features for Detecting Silicone Mask based Face Presentation Attack

摘要: Face recognition algorithms are generally vulnerable towards presentation attacks ranging from cost-effective ways such as print and replay to sophisticated mediums silicone masks. Carefully designed masks have real-life face texture once wore can exhibit facial motions; thereby making them challenging detect. In the literature, while several been developed for detecting based attacks, limited work has done mask-based attack. this research, we propose a computationally efficient solution by utilizing power of CNN filters, encoding mask attacks. The proposed framework operates on principle binarizing image region after convolving with filters learned via operations. On silicon attack database (SMAD), feature descriptor shows 3.8% lower error rate than state-of-the-art algorithms.