Abstract: As traditional privacy-preserving technology can't be directly applied to the social network data of higher dimension, to solve published graph data for social network facing the issues of privacy disclosure, a k-isomorphism privacy protection algorithm is proposed.By the original graph data is divided into k sub-graphs effectively, in order to reduce the cost of anonymity, the number of edges added edges approximately equal to the deleted, and ensure the release of the graph data is the k isomorphic, which effectively prevents the attacker based on a structural background knowledge attack.Real data set by experiment results show that the algorithm has high validity and can reduce the loss of information, also improve the quality of anonymity.