Avatar Facial Biometric Authentication Using Wavelet Local Binary Patterns

Virtual worlds (e.g., Second Life) are populated by different types of people, businesses, and organizations. Users of virtual worlds, either individuals or organizations, might abuse the flexibility and adaptability offered by virtual environment by engaging in criminal activities. Even terrorist organizations have been active in virtual worlds. These organizations recently have used the virtual worlds for recruitment and to train their new members in an environment that is very similar to the real one. Since avatars are not just virtual creations as they have a great social and psychological correspondence with their creators, applying biometric techniques on avatars can give the law enforcement agencies and security experts the ability to identify who the actual users behind these avatars are. There is a mounting pressure to have techniques for verifying the real identities of the inhabitants of virtual worlds to secure cyberworld from incessant criminal activities (e.g., verbal harassment, fraud, money laundering, data or identity theft). In order to reduce the gap between our ability to recognizing human faces and avatar faces and to develop reliable tools for protecting virtual environments, we will discuss in this chapter how we can use different versions of local binary pattern (LBP) operators (traditional LBP, multi-scale LBP and hierarchical multi-scale LBP) to recognize avatar faces from two different virtual worlds (Second Life and Entropia Universe). This chapter includes a definition of discrete wavelet transform from a face recognition research perspective, a summary of previous work done on this topic, characteristics of the datasets used in the experiments as well as some suggestions for future work.