Mediated Secure Multi-Party Protocols for Collaborative Filtering

Erez Shmueli, Tamir Tassa

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

Recommender systems have become extremely common in recent years and are utilized in a variety of domains such as movies, music, news, products, restaurants, and so on. While a typical recommender system bases its recommendations solely on users' preference data collected by the system itself, the quality of recommendations can significantly be improved if several recommender systems (or vendors) share their data. However, such data sharing poses significant privacy and security challenges, both to the vendors and the users. In this article, we propose secure protocols for distributed item-based Collaborative Filtering. Our protocols allow to compute both the predicted ratings of items and their predicted rankings without compromising privacy nor predictions' accuracy. Unlike previous solutions in which the secure protocols are executed solely by the vendors, our protocols assume the existence of a mediator that performs intermediate computations on encrypted data supplied by the vendors. Such a mediated setting is advantageous over the non-mediated one since it enables each vendor to communicate solely with the mediator. This yields reduced communication costs, and it allows each vendor to issue recommendations to its clients without being dependent on the availability and willingness of the other vendors to collaborate.

Original languageEnglish
Article number15
JournalACM Transactions on Intelligent Systems and Technology
Volume11
Issue number2
DOIs
StatePublished - 24 Feb 2020
Externally publishedYes

Keywords

  • Item-based collaborative filtering
  • distributed computing
  • privacy

Fingerprint

Dive into the research topics of 'Mediated Secure Multi-Party Protocols for Collaborative Filtering'. Together they form a unique fingerprint.

Cite this