The growing popularity of mobile and wearable devices with builtin cameras, the bright prospect of camera related applications such as augmented reality and lifelogging system, the increased ease of taking and sharing photos, along with advances in computer vision techniques, have greatly facilitated peoples lives in many aspects, but inevitably raised peoples concerns about visual privacy at the same time.

Motivated by the finding that peoples privacy concerns are influenced by the context, in this thesis, we propose Cardea, a context–aware and interactive visual privacy control framework that enforces privacy policies according to peoples privacy preferences. The framework provides people with finegrained visual privacy control using: i) personal privacy profiles, with which p...[ Read more ]

The growing popularity of mobile and wearable devices with builtin cameras, the bright prospect of camera related applications such as augmented reality and lifelogging system, the increased ease of taking and sharing photos, along with advances in computer vision techniques, have greatly facilitated peoples lives in many aspects, but inevitably raised peoples concerns about visual privacy at the same time.

Motivated by the finding that peoples privacy concerns are influenced by the context, in this thesis, we propose Cardea, a context–aware and interactive visual privacy control framework that enforces privacy policies according to peoples privacy preferences. The framework provides people with finegrained visual privacy control using: i) personal privacy profiles, with which people can define their context–dependent privacy preferences; ii) natural visual indicators: face features, for devices to automatically locate individuals who request privacy protection; iii) hand gestures, for people to temporarily update and flexibly inform cameras of their privacy preferences. Benefited from recent progresses in face and object recognition, Cardea offers a way for context–dependent privacy control in a natural and flexible manner, which differs from tag and marker based systems. We design and implement the framework consisting of Android client app and cloud control server, with convolutional neural networks as core of the image processing module. Our evaluation results confirm such framework is practical and effective, showing promising future for context–aware visual privacy control on mobile and wearable devices.