Towards user-oriented privacy for recommender system data: A personalization-based approach to gender obfuscation for user profiles
作者：
Highlights：
• We introduce PerBlur, an approach that obfuscates recommender system data.
• PerBlur uses personalized blurring to block inference of users’ gender.
• We describe the user-oriented privacy paradigm in which PerBlur is formulated.
• We propose an evaluation procedure for obfuscated recommender system data.
• PerBlur is demonstrated to be capable of maintaining recommender system performance.
• We show the potential of obfuscation to improve fairness and diversity.
摘要
•We introduce PerBlur, an approach that obfuscates recommender system data.•PerBlur uses personalized blurring to block inference of users’ gender.•We describe the user-oriented privacy paradigm in which PerBlur is formulated.•We propose an evaluation procedure for obfuscated recommender system data.•PerBlur is demonstrated to be capable of maintaining recommender system performance.•We show the potential of obfuscation to improve fairness and diversity.
论文关键词：Top-N recommendation,Obfuscation,Gender inference,Evaluation,Privacy,Fairness,Diversity
论文评审过程：Received 30 November 2020, Revised 30 July 2021, Accepted 4 August 2021, Available online 14 September 2021, Version of Record 14 September 2021.
论文官网地址：https://doi.org/10.1016/j.ipm.2021.102722
