Conference Paper

Towards Efficient Privacy-preserving Collaborative Recommender Systems.

Heinz Sch., Carnegie Mellon Univ., Pittsburgh, PA
DOI: 10.1109/GRC.2008.4664769 In proceeding of: The 2008 IEEE International Conference on Granular Computing, GrC 2008, Hangzhou, China, 26-28 August 2008
Source: DBLP

ABSTRACT Recommender systems use various types of information to help customers find products of personalized interest. To increase the usefulness of recommender systems in certain circumstances, it could be desirable to merge recommender system databases between companies, thus expanding the data pool. This can lead to privacy disclosure hazards that this paper addresses by constructing an efficient privacy-preserving collaborative recommender system based on the scalar product protocol.

0 Bookmarks
 · 
66 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: Protocols for secure multi-party computation allow participants to share a computation while each party learns only what can be inferred from their own inputs and the output of the computation. However, the execution time of a secure protocol may be too high therefore it is not practical unless some tradeoffs are made between data access and confidentiality. This paper aims to provide some empirical basis for making such tradeoffs in computing exponentiation. We have designed exponentiation protocols for secure two-party computation using scalar products as the basic building blocks. A detailed performance evaluation was carried out by taking advantage of the compositional nature of our protocols. We have come up with a time function which provides good prediction of the execution time of the proposed exponentiation protocols based on the execution time of scalar products. Using the time function, we have obtained several interesting tradeoffs between execution time and privacy. In particular, compromising some private information enables a reduction in the execution time from years, if not centuries, to days or even minutes. Based on our results, we argue that there are indeed reasonable tradeoffs between privacy and execution time. Furthermore, our study indicates that a system intelligently offering users possible tradeoff options will make secure multi-party computation a more attractive approach to enhancing privacy in practice.
    2010 13th IEEE International Conference on Computational Science and Engineering. 08/2009; 3:182-188.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Protocols for secure multi-party computation allow par-ticipants to share a computation while each party learns only what can be inferred from their own inputs and the output of the computation. However, the execution time of a secure protocol may be too high so that it is not practical unless some tradeoffs being made between data access and confidentiality. In this technical report, we propose a set of information theoretically secure protocols based on scalar product protocol and aim to provide some empirical basis for making such tradeoffs in computing exponentiation. A detailed performance evaluation was carried out by taking advantage of the compositional nature of our protocols. We come up with a time function which provides good predic-tion of the execution time of the proposed exponentiation protocols based on the execution time of scalar products. Using the time function, we obtain several interesting trade-offs between execution time and privacy. In particular, com-promising some private information enables a reduction in the execution time from years, if not centuries, to days or even minutes. Based on our results, we argue that there are indeed reasonable tradeoffs between privacy and execution time. Furthermore, our study indicates that a system intel-ligently offering users possible tradeoff options will make secure multi-party computation a more attractive approach to enhance privacy in practice.
    01/2010;

Full-text (2 Sources)

View
4 Downloads
Available from
Jun 5, 2014