The rapid development and adoption of online social networks have allowed us to establish communities and create online personas that reflect our personal interests and orientations (political preferences, ethnicity, gender orientation, etc). However, due to the rise of machine learning and deep learning techniques, it is possible for malicious individuals to utilize social media posts, likes, and interactions to predict certain sensitive attributes about a user, such as fine-grain location inferences up to 70% accuracy.
Aegis is an artificially intelligent application, which resides locally on a user’s machine providing tools for the user to obfuscate their location from adversaries while preserving their online persona. My project within Aegis focuses on creating the data structures necessary to efficiently and effectively provide obfuscation mechanisms for a user’s location while ensuring their online persona stays constant. The privacy model and system is then built on top of these data structures which allow a user to reach k-indistinguishability for location.