Privacy Preserving Collaborative Filtering using Biclustering in Ubiquitous Computing Environments

Privacy concerns are a major hurdle in the success of personalized services in ubiquitous computing environments. Personalized recommendations are usually served using Collaborative Filtering techniques. In this paper, we propose a framework for privacy preserving collaborative filtering in ubiquitous computing environments. The proposed framework is based on a biclustering algorithm which employs bipartite graph crossing minimization heuristic as an efficient replacement of Spectral Partitioning techniques. Distributed biclustering eliminates the requirement of trusted server. Privacy is ensured by Secure multiparty computations that are carried out using an additively homomorphic cryptosystem. The proposed framework is secure under a malicious adversary model where nearly half of the users are allowed to collude. Theoretical complexity analysis shows that the proposed system is scalable in terms of computation and communication complexity.