Privacy Preserving Biometrics

Slides (pdf)  Size: 11.9MB

Abstract

In many face recognition applications, it is necessary to store the raw face images of individuals in a database. Recent research has demonstrated that attributes such as age, gender, race, body-mass index, etc. can be extracted from these images using attribute classifiers. Automated extraction of such information can breach personal privacy, especially if the information is used for purposes beyond what was stated at the time of collecting the data. The gleaned information can also be used for profiling individuals. While image encryption schemes can impart security to face images in a database, it is still possible for the administrators of a database to extract attributes when the image is decrypted at the time of recognition. In this lecture, we will discuss the use of image perturbation schemes to enhance the privacy of raw face images. In particular, we will describe semi-adversarial neural networks, or SANs, that can be used to modify face images in a database. SANs produce images that are semi-adversarial, i.e., the performance of certain attribute classifiers is negatively impacted, while the performance of face matchers is not adversely impacted. This facilitates controllable privacy, where attribute extraction is suppressed while biometric utility is retained. We will present experimental results based on multiple face datasets, attribute classifiers and face matchers in order to convey the efficacy of SANs in imparting privacy to face images.

Biography

Arun Ross is a Professor in the Department of Computer Science and Engineering at Michigan State University, and is the Director of the Integrated Pattern Recognition and Biometrics (iPRoBe) Lab. He also serves as the Site Director of the NSF Center for Identification Technology Research (CITeR). Ross conducts research on the topic of biometrics, privacy, computer vision and pattern recognition. He is a recipient of the JK Aggarwal Prize and the Young Biometrics Investigator Award from the International Association of Pattern Recognition for his contributions to the field of Pattern Recognition and Biometrics. He was designated a Kavli Fellow by the US National Academy of Sciences by virtue of his presentation at the 2006 Kavli Frontiers of Science Symposia. Ross was an invited panelist at a counter-terrorism event organized by the United Nations Counter-Terrorism Committee (CTC) at the UN Headquarters in 2013. He has advocated for the responsible use of biometrics in multiple forums including the NATO Advanced Research Workshop on Identity and Security in Switzerland in 2018. He is a recipient of the NSF CAREER Award, the 2005 Biennial Pattern Recognition Journal Best Paper Award and the Five Year Highly Cited BTAS 2009 Paper Award. Ross is the co-author of the monograph “Handbook of Multibiometrics” and the textbook “Introduction to Biometrics”.

Arun Ross
Michigan State University, US