In collaborative filtering approaches, recommendations are inferred from user data. A large volume and a high data quality is essential for an accurate and precise recommender system. As consequence, companies are collecting large amounts of personal user data. Such data is often highly sensitive and ignoring users’ privacy concerns is no option. Companies address these concerns with several risk reduction strategies, but none of them is able to guarantee cryptographic secureness. To close that gap, we propose a novel recommender system using the advantages of blockchain-supported secure multiparty computation. A potential costumer is able to allow a company to apply a recommendation algorithm without disclosing her personal data. Expected benefits are a reduction of fraud and misuse and a higher willingness to share personal data.
We intend to develop a first prototype of such a recommender system. Besides the technical challenges, one research focus lies on the impact of the system on customers’ privacy-related behavior.