Protecting personal identification in video

Abstract

In this chapter, we present some studies on protecting personal identification in video. First, we discuss and evaluate automatic face masking techniques for obscuring human faces in video. Second, a user study is presented to reveal that face-masked video can be attacked using pair-wise constraints. Next, we propose an algorithm to show that this type of pair-wise constraint attack can be implemented using state-of-the-art machine learning approaches. Finally, a new obscuring approach is proposed to avoid the pair-wise constraint attack. The proposed approach protects people's identity by obscuring the texture information of the entire body. © 2009 Springer London.