Enhancing Social Matrix Factorization with Privacy
2013 (English)In: Proceedings of the ACM Symposium on Applied Computing, 2013, 277-278 p.Conference paper, Poster (Refereed)
Within the course of this manuscript we present a privacy-preserving collaborative filtering recommender system whichaims at alleviating the concern with privacy of user pro-files within the context of sparse social trust data. Whileproblem of sparsity in social trust is often addressed by tak-ing similarity driven trust measures through a probabilisticmatrix factorization technique, we address the issue of pri-vacy by proposing a dynamic privacy inference model. Theprivacy inference model exploits the underlying inter-entitytrust information in order to build a personalized privacyperspective for each individual within the social network.This is followed by our evaluation of the proposed solutionby adopting an off-the-shelf collaborative filtering recom-mender library, in order to generate predictions using thispersonalized view.
Place, publisher, year, edition, pages
2013. 277-278 p.
Matrix factorization, Privacy, Privacy inference, Recommender systems, Social network, Trust
IdentifiersURN: urn:nbn:se:kth:diva-118475DOI: 10.1145/2480362.2480421ScopusID: 2-s2.0-84877981601ISBN: 978-145031656-9OAI: oai:DiVA.org:kth-118475DiVA: diva2:606436
28th Annual ACM Symposium on Applied Computing, SAC 2013; Coimbra, Portugal, 18 March through 22 March 2013
QC 201309032013-02-192013-02-192014-01-24Bibliographically approved