In order to protect the normal operation of social networks from the influence of malicious communication, we established a system to help evaluate information and users. On one hand, based on behavior recognition technology, the information model uses the specific information's operation data as one of the information's special features to analyze its reliability. On the other hand, to minimize the impacts of some identity's malicious use, we use a two-dimensional algorithm, identifying the users' abnormal behaviors through comparison, as our user model. Our system can effectively reduce malicious communication by identifying malicious information and users in social networks.