Client-Server Framework for Securely Outsourcing Computations

In recommendation systems, service providers can recommend products to their customers based on personal data from many users. To avoid leakage of these personal data, service providers use secret sharing. This enables them to compute the recommendation, while remaining oblivious to the contents. Only the user requesting the output will be able to combine all shares of the recommendation and learn the result. Although similar secure recommendations systems have been developed before, they were either less secure, or less efficient. This framework is the first one that uses fast cryptographic techniques within a ‘malicious security model’. By precomputing a lot of input independent data, the two servers were able to compute a recommendation