位置信息共享下面向智能服务推荐的隐私保护方法研究

In the future, location-based service recommendation in location information sharing with high accuracy, large scale and personality heavily depends on the credible data privacy-preserving and publishing. However, current solutions are suffering from the unawareness of user behavior, complex context information and diversified service characteristics, and the lack of collaboration between privacy-preserving and service recommendation. These drawbacks not only lead to low data availability of central server but also make it difficult to ensure the recommendation accuracy of service and the quality of experience of users. As a result, the demand of smart service recommendation system in location information sharing cannot be fulfilled. To solve these problems, this project aims to study the privacy-preserving methods for smart service recommendation system. The main contents include: (1) Establishing user preference model based on knowledge graph, in order to accurately extract the characteristics of interest and preference for users. (2) Researching on context and preference aware trajectory privacy-preserving method. Quantizing the strength of privacy-preserving by making use of the theories of information entropy, game theory, etc. (3) Researching on semantic features fused differential privacy trajectory data publishing method, in order to solve the collaborative problem between privacy-preserving and service recommendation, and increase the data utility. (4) Building the experimental platform of smart service recommendation supported by privacy-preserving, so as to evaluate the feasibility of proposed theories and methods. Based on the above researches, smart service recommendation in location information sharing can be realized, the data security can be significantly improved, and the high-quality location-based service can be guaranteed.

未来位置共享场景下高精度、大规模、个性化的位置服务推荐依赖于可信的数据隐私保护与发布。然而传统解决方案由于无法立体感知用户行为、复杂情景信息和多样化服务特性，且隐私保护与服务推荐缺乏协同，造成中心服务器所获数据的可用性低，服务推荐精度及用户体验质量得不到有效保障，难以适应位置信息共享下智能服务推荐系统的发展需求。为此，本项目拟开展面向智能服务推荐的隐私保护方法研究。主要研究内容包括：（1）建立基于知识图谱的用户行为模型，精确提取用户的兴趣及偏好特征；（2）研究情景与偏好感知的轨迹隐私保护方法，运用信息熵、博弈论等理论量化隐私保护强度；（3）研究融合语义特征的差分隐私轨迹数据发布方法，解决隐私保护与服务推荐协同问题，提高数据的可用性；（4）搭建支持隐私保护的智能服务推荐实验平台，评估所提理论与方法的可行性。通过上述研究，实现位置信息共享下智能服务推荐，大幅提高数据安全性并保障高质量位置服务。

未来位置共享场景下高精度、大规模、个性化的位置服务推荐依赖于可信的数据隐私保护与发布，传统解决方案由于无法立体感知用户行为、复杂情景信息和多样化服务特性，且隐私保护与服务推荐缺乏协同，造成中心服务器所获数据的可用性低，服务推荐精度及用户体验质量得不到有效保障，难以适应位置信息共享下智能服务推荐系统的发展需求。针对以上问题，本项目开展了面向智能服务推荐的隐私保护方法研究，主要包括：（1）建立基于知识图谱的用户行为模型，精确提取用户的兴趣及偏好特征；（2）研究情景与偏好感知的轨迹隐私保护方法，运用信息熵、博弈论等理论量化隐私保护强度；（3）研究融合语义特征的差分隐私轨迹数据发布方法，解决隐私保护与服务推荐协同问题，提高数据的可用性；（4）搭建支持隐私保护的智能服务推荐实验平台，评估所提理论与方法的可行性。上述研究工作为高质量的位置隐私保护提供了一整套行之有效的解决方案。经过3年的研究工作，项目组共计发表论文13篇，其中SCI10篇，EI2篇，核心1篇，超额完成了任务。申请专利14项，其中获得授权5项，为研究成果的推广奠定了基础。项目在研期间，课题组培养了硕士研究生3名，邀请了2名国内学者来实验室进行交流，并与航天时空信息技术（河南）有限公司合作完成横向项目，实现成果转化。