Smart eyewear devices, such as augmented reality displays, increasingly integrate eye tracking. Unsurprisingly, as any (front-facing) body-worn camera, the scene camera poses a serious privacy risk. It may not only impair bystander privacy, but also record sensitive personal information, such as login credentials, banking information, or text messages. Physical shutters are a compelling option to protect privacy. Yet, with the camera shutter closed, visual scene knowledge is not available, and the camera can't automatically reactivate.
As a solution, we propose PrivacEye, a smart glasses prototype that infers the right moment for camera re-activation from the user’s eye movements. To close the shutter in privacy-sensitive situations, the method uses a deep representation of the first-person video combined with rich features that encode users' eye movements. To open the shutter without visual input, PrivacEye detects changes in users' eye movements alone to gauge changes in the ''privacy level'' of the current situation. We evaluate our method on a first-person video dataset recorded in daily life situations of 17 participants, annotated by themselves for privacy sensitivity, and show that our method is effective in preserving privacy in this challenging setting.