Examining users' attitude towards privacy preserving collaborative filtering

Shlomo Berkovsky, Nikita Borisov, Yaniv Eytani, Tsvi Kuflik, Francesco Ricci

Research output: Chapter in Book/Report/Conference proceedingConference proceeding contributionpeer-review


Privacy hazard to Web-based information services represents an important obstacle to the growth and diffusion of the personalized services. Data obfuscation methods were proposed for enhancing the users’ privacy in recommender systems based on collaborative filtering. Data obfuscation can provide statistically measurable privacy gains. However, these are measured using metrics that may not be necessarily intuitively understandable by end user,
such as conditional entropy. In fact, it could happen that the users are unaware, misunderstand how their privacy is being preserved or do not feel comfortable with such methods. Thus, these may not reflect in the users’ actual personal sense of privacy. In this work we provide an exploratory study to examine correlation between different data obfuscation methods and their effect on the subjective sense of privacy of users. We analyze users’ opinion about the impact of data obfuscation on different types of users’ rating values and generally on their sense of privacy.
Original languageEnglish
Title of host publicationData Mining for User Modeling
Subtitle of host publicationOn-line Proceedings of Workshop held at the International Conference on User Modeling, UM 2007
Number of pages7
Publication statusPublished - 2007
Externally publishedYes
EventInternational Conference on User Modeling (11th : 2007) - Corfu, Greece
Duration: 25 Jun 200729 Jun 2007


ConferenceInternational Conference on User Modeling (11th : 2007)
CityCorfu, Greece


Dive into the research topics of 'Examining users' attitude towards privacy preserving collaborative filtering'. Together they form a unique fingerprint.

Cite this