针对社交网络边权重的差分隐私保护

Differential privacy protection for social network edge weights

针对社交网络边权重隐私保护中的弱保护和最短路径不可分析问题,提出一种满足差分隐私保护模型的边权重保护策略。将社交网络划分为全次图、缺次图、零次图,设计扰动方案及查询函数,对不同图进行查询获取其边权重并按扰动方案对不同的边权重添加不同的Laplace噪声,实现抵御攻击者拥有最大背景知识的攻击的边权重隐私保护,保证一组节点的最短路径不变,且其长度与原路径长度相近。该策略有强保护性及最短路径可分析性,从理论上验证了算法的可行性,通过实验验证了算法的正确性。

Aiming at the weak protection and unanalysable shortest path in the edge weight privacy protection of social network,an edge weight protection strategy based on differential privacy model was proposed.The social network was divided into whole-timesgraph,sub-times-graph and zero-times-graph.By designing the perturbation scheme and the query function,the edge weights of the different graphs were got and the different Laplace noise was added to different edge weights according to the perturbation scheme.The edge weight privacy protection against the attacker with the largest background knowledge was realized,and the shortest path of a group of nodes was kept unchanged and its length was close to the original path length.The strategy has strong protection ability and analysable shortest path.The feasibility of the algorithm is proved theoretically,and the correctness of the algorithm is verified by experiments.