k-CoRating: filling up data to obtain privacy and utility

摘要: For datasets in Collaborative Filtering (CF) recommendations, even if the identifier is deleted and some trivial perturbation operations are applied to ratings before they released, there research results claiming that adversary could discriminate individual's identity with a little bit of information. In this paper, we propose k-coRating, novel privacy-preserving model, retain data privacy by replacing null "well-predicted" scores. They do not only mask original such k-anonymity-like preserved, but also enhance utility (measured prediction accuracy paper), which shows traditional assumption two goals conflict necessarily correct. We show optimal k-coRated mapping an NP-hard problem design naive efficient algorithm achieve k-coRating. All claims verified experimental results.