Focusing on the issue that the current related research about social network do not consider subsets for neighborhood's privacy preserving, and the specific properties of neighborhood subsets also lead individual privacy disclosure, a new (θ, k)-anonymous model was proposed. According to the k-isomorphism ideology, the model removed labels of neighborhood subsets which needed to be protected in social network, made use of neighborhood component coding technique and the method of node refining to process nodes in candidate set and their neighborhood information, then completed the operation of specific subsets isomorphism with considering the sensitive attribute distribution. Ultimately, the model satisfies that each node in neighborhood subset meets neighborhood isomorphism with at least k-1 nodes, as well the model requires the difference between the attribute distribution of each node in the neighborhood subset and the throughout subsets is not bigger than θ. The experimental results show that, (θ, k)-anonymous model can reduce the anonymization cost and maximize the utility of the data.