Fantômas: Understanding Face Anonymization Reversibility
This work addresses privacy risks in face anonymization for users by exposing flaws in existing methods, though it is incremental as it builds on prior findings of reversibility.
The paper investigates the reversibility of face anonymization techniques, finding that 11 out of 15 tested methods are at least partially reversible, with reconstruction and inversion identified as key processes enabling reversal.
Face images are a rich source of information that can be used to identify individuals and infer private information about them. To mitigate this privacy risk, anonymizations employ transformations on clear images to obfuscate sensitive information, all while retaining some utility. Albeit published with impressive claims, they sometimes are not evaluated with convincing methodology. Reversing anonymized images to resemble their real input -- and even be identified by face recognition approaches -- represents the strongest indicator for flawed anonymization. Some recent results indeed indicate that this is possible for some approaches. It is, however, not well understood, which approaches are reversible, and why. In this paper, we provide an exhaustive investigation in the phenomenon of face anonymization reversibility. Among other things, we find that 11 out of 15 tested face anonymizations are at least partially reversible and highlight how both reconstruction and inversion are the underlying processes that make reversal possible.