Generating Private Recommendations in a Social Trust Network Z. Erkin * , T. Veugen *† and R. L. Lagendijk * * Information Security and Privacy Lab Delft University of Technology, 2628 CD, Delft, The Netherlands † TNO, P.O. Box 5050, 2600 GB, Delft, The Netherlands Email: {z.erkin, p.m.j.veugen, r.l.lagendijk}@tudelft.nl Abstract—Recommender systems have become increasingly important in e-commerce as they can guide customers with finding personalized services and products. A variant of rec- ommender systems that generates recommendations from a set of trusted people is recently getting more attention in social networks. However, people are concerned about their privacy as the information revealed in recommender systems, particularly in social networks, can be misused easily. A way to eliminate the privacy risks is to make the privacy-sensitive data inaccessible by means of encryption. While the private data is inaccessible to any outsiders, the same functionality of the system can be achieved by processing the encrypted data. Unfortunately, the efficiency of processing encrypted data constitutes a big challenge. In this paper, we present a privacy-enhanced recommender system in a social trust network, which is designed to be highly efficient. The cryptographic protocol for generating recommendations is based on homomorphic encryption and secure multi-party computation techniques. The additional overhead with regard to computation and communication is minimized by packing data. The experimental results show that the performance of our proposal is promising to be deployed in real world. Index Terms—Social trust networks, recommender systems, privacy, homomorphic encryption, secure multi-party computa- tion. I. I NTRODUCTION Recommender systems have become increasingly important for e-commerce as these systems help customers in decision making for a service or a product by providing personal- ized recommendations. The recommender systems based on collaborative filtering techniques are being used widely in numerous e-commerce applications like Amazon, eBay and youtube. However, in certain online applications, particularly those that also involve social interaction, people are interested in the recommendations of a particular group which consists of people with a certain trust relation rather than a number of similar people in the network. These networks, so called Social Trust Networks (STNs), are attracting more attention recently as they can be used for more accurate recommenda- tion generation [12], [16], [23]. In an STN, a network graph or a matrix depicting the trust relation among the users of a recommender system is created. As an example Fig. 1 consists of 5 nodes, each representing a user. The link between nodes represents the trust relation: it either exists or not. Notice that the graph is symmetric, meaning that user A trusts user B and vice versa. However, the graph can also be directional. Given a graph or matrix representation of a trust network, recommendations can be easily generated by averaging the ratings of the trusted users. 1 2 3 4 5 1 0 1 0 1 1 2 1 0 1 1 0 3 0 1 0 1 0 4 1 1 1 0 0 5 1 0 0 0 0 Fig. 1. Trust relation graph and the corresponding matrix. Recommender systems in general, including STNs, possess high privacy risks for the users [21]. The privacy concerns arise because the recommendations are generated by a central entity, the service provider, and this entity has access to privacy- sensitive information, which can be easily used to identify and track any individual. Moreover, the private information on the user of recommender systems can be re-purposed or transferred to third parties without the knowing of the users. The risk is even higher in the case of STNs as these systems have a social network part integrated where people tend to reveal personal information. For example, any misuse in a medical application where patients are suggested contents or encouraged to make contact with other patients with the same disease as in www.patientslikeme.com will result in considerable damage on privacy. In this paper, we present an efficient privacy-preserving rec- ommender system for STNs that hides the private information from the service provider. We achieve this goal by providing only encrypted data to the service provider, which cannot access to the private data directly but has cryptographic tools to process the encryptions to generate the recommendations. The cryptographic tools designed for this purpose are based on Homomorphic Encryption (HE) and Multi-Party Computa- tion (MPC) techniques. The privacy-preserving recommender system requires more resources compared to its plain version due to operating in the encrypted domain. Therefore, we focus on efficiency and present a fine-tuned cryptographic protocol with low storage, bandwidth and computation requirements.