MS Project Defense -- Ishani Gupta

Thursday, May 17, 2018 - 4:00pm
HFH 1132
Privacy Model to Hide your Location from an Online Persona Aware Adversary
Amr El Abbadi (Chair), Ambuj Singh

A social media’s user develops an online persona (political preferences, ethnicity, gender orientation, religious preferences) that reflects her/his overall interests, activism, and diverse orientations. Recent works show that it is possible to infer much more information from the user’s social media account than intended to be shared by the user. One of these attributes which can be inferred and misused is user’s location. With the advancement in natural language processing, only with the contextual analysis of the social media streams, the location of a user can be inferred with up to 70 % accuracy.

Aegis is a tool that provides the user with the power to control her/his "unshared yet revealed” location information. Aegis is  an artificially intelligent cyborg which acts as a personal bodyguard for the social media.  It is truly personal because it is locally deployed at user’s machine and does not require any centralized information sharing. My project within Aegis focuses on building the core privacy model of obfuscation to achieve k-location indistinguishability without affecting the user’s online persona. K-location indistinguishability is achieved by suggesting topics to user which align with her/his online persona but also aim to equalise the adversary’s confidence in user’s location among k different locations.



Everyone welcome!