Chapter 11 PRIVACY PROTECTION IN COLLABORATIVE FILTERING BY ENCRYPTED COMPUTATION

We present a method to protect users’ privacy in collaborative filtering by performing the computations on encrypted data. We focus on the commonly-used memory-based approach, and show that the two main steps in collaborative filtering, being the determination of similarities and the prediction of ratings, can be performed on encrypted profiles. We discuss both user-based and item-based collaborative filtering, and for a number of variants of the similarity measures and prediction formulas described in literature, we show how they can be computed using encrypted data only. Although we consider collaborative filtering in this chapter, the techniques of comparing profiles using encrypted data only is useful in a much wider range of applications.