Ulixes: Facial Recognition Privacy with Adversarial Machine Learning
This addresses privacy concerns for users of online services with photo management, such as social media platforms, by providing a robust defense against facial recognition tools, even in black-box settings and against adversarially trained models.
The paper tackles the privacy problem in facial recognition by proposing Ulixes, a method to generate facial noise masks that create adversarial examples, preventing identifiable user clusters in facial encoders, and demonstrates its effectiveness by showing various classification and clustering methods cannot reliably label these examples.
Facial recognition tools are becoming exceptionally accurate in identifying people from images. However, this comes at the cost of privacy for users of online services with photo management (e.g. social media platforms). Particularly troubling is the ability to leverage unsupervised learning to recognize faces even when the user has not labeled their images. In this paper we propose Ulixes, a strategy to generate visually non-invasive facial noise masks that yield adversarial examples, preventing the formation of identifiable user clusters in the embedding space of facial encoders. This is applicable even when a user is unmasked and labeled images are available online. We demonstrate the effectiveness of Ulixes by showing that various classification and clustering methods cannot reliably label the adversarial examples we generate. We also study the effects of Ulixes in various black-box settings and compare it to the current state of the art in adversarial machine learning. Finally, we challenge the effectiveness of Ulixes against adversarially trained models and show that it is robust to countermeasures.