基于半监督学习的社交网络用户属性预测

研究如何利用社交关系推测用户的隐藏属性(私隐信息)，采用基于图的半监督学习方法推测用户属性。为了提高预测的准确率，提出利用属性聚集度评价属性推测的难易程度，并依用户节点标记的不同，设计不同的权重公式计算用户之间的关系强度。以“人人网”数据作为实验数据，对用户的兴趣与毕业学校进行预测，验证了方法的有效性。

Privacy-perserving scheme for social networks

The security and privacy issues in SNS were studied and a privacy-preserving scheme PPSNS was proposed. PPSNS utilizes attribute-based encryption, allowing the SNS user to set up an enforcement of fine-grained access control upon the data he owns, thus the potential threats from unauthorized parties or even the SNS provider itself could be avoided. A token system in PPSNS is included to address the challenging issue of efficient attribute revocation. In addi-tion, the users in PPSNS don't have to manage as much information as they do in other encryption-based solutions, achieving a much lower cost in the client side. Analyses show that PPSNS is secure, and gives a better performance in computing and storage costs compared to most related works.