服务推荐中基于局部敏感哈希的隐私保护方法研究

Service recommendation is one of the most effective approaches to quickly extract insightful information from big data. The characteristics of big data, such as multi-source property and data diversity, require that a recommender system can quickly integrate the data distributed across multiple parties so as to make comprehensive and accurate recommendation decisions. In particular, to protect business secrets and obey laws, preserving user privacy during the abovementioned integration process is a very important but challenging requirement in practice. Locality-Sensitive Hashing (LSH) is one of the most effective techniques to achieve the abovementioned privacy-preservation goal. However, due to the built-in drawbacks of LSH and the diversity of big data, existing LSH-based service recommendation methods still face many challenges that need to be addressed urgently. In view of these challenges, in this project, we first theoretically analyze the intrinsic drawbacks of LSH when it is applied to service recommendation, and systematically establish an LSH-based privacy-preserving model for service recommendation. Then, a series of novel LSH-based service recommendation algorithms will be designed and developed by considering the requirements of service recommendation and privacy preservation in the big data context. The expected research outcome of this project will fundamentally and systematically build up the theoretical framework of privacy preservation in LSH-based recommender systems, and promote seamless and secure fusion of recommender systems and big data. Furthermore, the developed light-weight service recommendation techniques will address the information explosion challenges incurred by big data.

服务推荐是从大数据中快速提取价值信息的有效途径之一。大数据的特点如多源性及多样性，要求推荐系统能对分布于多个提供方的数据进行快速集成，以便提供全面而准确的推荐决策。特别地，为了商业机密和遵守法律，保护用户隐私在上述集成过程是一个非常重要但具有挑战性的现实要求。局部敏感哈希（LSH）是实现上述隐私保护目标的有效技术之一。然而，由于LSH的固有缺陷及大数据的多样性，目前基于LSH的服务推荐方法仍然面临许多亟待解决的挑战。鉴于此，本课题首先从理论层面分析LSH对于服务推荐的固有缺陷，并系统地建立基于LSH的隐私保护模型；然后结合大数据环境下的服务推荐及隐私保护需求，提出并实现一系列新颖的基于LSH的服务推荐算法。课题的预期研究结果，将会基础性、系统性地建立起LSH推荐系统中的隐私保护理论体系，从而促进推荐系统和大数据的紧密安全融合，进而通过所提出的轻量化服务推荐技术，来应对大数据导致的信息爆炸。

服务推荐是从大数据中快速提取价值信息的有效途径之一。大数据的特点如多源性及多样性，要求推荐系统能对分布于多个提供方的数据进行快速集成，以便提供全面而准确的推荐决策。特别地，为了商业机密和遵守法律，保护用户隐私在上述集成过程是一个非常重要但具有挑战性的现实要求。局部敏感哈希（LSH）是实现上述隐私保护目标的有效技术之一。然而，由于LSH的固有缺陷及大数据的多样性，目前基于LSH的服务推荐方法仍然面临许多亟待解决的挑战。鉴于此，本课题首先从理论层面分析了LSH对于服务推荐的固有缺陷，并系统地建立了基于LSH的隐私保护模型；然后结合大数据环境下的服务推荐及隐私保护需求，提出并实现了一系列新颖的基于LSH的服务推荐算法（如：基于时间、空间的LSH服务推荐算法、基于哈希增强的LSH服务推荐算法、考虑多样性、流行度、鲁棒性的LSH服务推荐算法，等等）。课题的研究结果，基础性、系统性地建立起了LSH推荐系统中的隐私保护理论体系，促进了推荐系统和大数据的紧密安全融合；进而通过项目提出的轻量化服务推荐技术，可以有效地应对大数据导致的信息爆炸。