FaceWard: Face Anonymization in Group Photos

Sharing photos on social media and messaging services often result in a vast amount of personal data being made public online. As a result, it has become increasingly vital to devise measures that ensure privacy protection, especially for people who want to maintain social boundaries by hiding their faces in group photos. In this paper, we propose FaceWard, an automatic system for face anonymization of people different from a target person. FaceWard is based on a pattern matching algorithm and supports different anonymization policies such as blur or smiley overlays. Taken together, FaceWard yields a very efficient solution, eliminating the need for computationally expensive training of complex machine learning models, thus offering a practical trade-off between prediction accuracy and data availability. FaceWard is publicly available as open source software.