高效可验证的隐私保护推荐系统

摘要 针对个性化推荐服务系统存在的隐私泄露问题，提出了一个高效可验证的隐私保护推荐系统，能在保护用户数据隐私的前提下，实现用户对云端计算出的推荐模型的正确性验证；利用脊回归实现对用户数据的拟合；利用Yao的混淆电路技术实现推荐模型的计算以及对模型的正确性验证.用户端和云端使用一种新的数据聚合算法AGG（Aggregation）来替换大多数已有工作中使用的公钥同态加密算法，减少了用户端和云端的计算开销，使得系统效率更高.给出了方案的安全性分析以及效率分析.
Abstract：To address the problem of privacy disclosure in traditional personalized recommendation systems, this paper proposes an efficient verifiable privacy-preserving recommendation system, which can provide user the way to verify the correctness of the resulting model of cloud computing under the premise of protecting user's data privacy. This paper uses ridge regression to find the best-fit linear curve of user's input data, and implements Yao's garbled circuit to realize the computation and the correctness verification of the recommendation model. The user and the cloud use a newly-devised privacy preserving data aggregation method named AGG (Aggregation) to replace public key homomorphic encryption used in most existing work, which can reduce the computational overhead of the user and the cloud, thus making the system more efficient. The security analysis and the efficiency analysis of the scheme are given at the end of the article.