PROJECT TITLE :

Privacy-Preserving Social Media Data Publishing for Personalized Ranking-Based Recommendation

ABSTRACT:

To assist users in finding relevant information, personalized recommendations are critical. To mine user preference, it frequently relies on a vast collection of user data, particularly users' online activity (e.g., tagging/rating/checking-in) on social media. However, exposing such user activity data exposes consumers to inference attacks, as private details (for example, gender) can typically be inferred from the data. We introduced PrivRank in this research as a customisable and continuous privacy-preserving social media data publishing platform that protects users from inference assaults while allowing personalized ranking-based suggestions. Its central concept is to continuously obfuscate user activity data in such a way that privacy leakage of user-specified private data is minimized within a given data distortion budget, which limits the ranking loss caused by the data obfuscation process in order to preserve the data's utility for enabling recommendations. Our methodology can efficiently provide effective and ongoing protection of user-specified private data while yet keeping the utility of the obfuscated data for personalized ranking-based recommendation, according to an empirical evaluation on both synthetic and real-world datasets. In all of the ranking-based recommendation use cases we evaluated, PrivRank offers both superior privacy protection and higher utility when compared to state-of-the-art techniques.