Utility-Preserving Biometric Information Anonymization

The use of biometrics such as fingerprints, voices, and images are becoming increasingly more ubiquitous through people's daily lives, in applications ranging from authentication, identification, to much more sophisticated analytics, thanks to the recent rapid advances in both the sensing hardware technologies and machine learning techniques. While providing improved user experiences and better business insights, the use of biometrics has raised serious privacy concerns, due to their intrinsic sensitive nature and the accompanying high risk of leaking personally identifiable and private information. In this paper, we propose a novel utility-preserving biometric anonymization framework, which provides a method to anonymize a biometric dataset without introducing artificial or external noise, with a process that retains features relevant for downstream machine learning-based analyses to extract interesting attributes that are valuable to relevant services, businesses, and research organizations. We carried out a thorough experimental evaluation using publicly available visual and vocal datasets. Results show that our proposed framework can achieve a high level of anonymization, while at the same time retain underlying data utility such that subsequent analyses on the anonymized biometric data could still be carried out to yield satisfactory accuracy.