To solve the attacks of sensitve edge identification using social individual neighborhood information as background knowledge inside the social networks, a random disturbance method based on spectrum constraint and sensitive area division was proposed. The main idea of this method was to divide the network into sensitive zone and non-sensitive zone. It compared the last disturbance spectrum size of the social network graph with the original social network graph, and chose the right edges to add, delete or convert based on the comparison results and spectrum constraints, thus improving the usability of social network data. This method can improve the degree of privacy protection by eliminating invalid disturbances. The experimental results show that it can protect structural characteristics of social network better.