A user's historical service data of a first time period and a second time period is obtained. Means and variances of a plurality of users' historical service data of the first time period and the second period are calculated. A risk model is generated based on at least one type of the plurality of users' historical behavioral data. The user's service data is obtained. At least one type of the user's behavior data is introduced to the risk model. A determination is made whether the at least one type of the user's behavior data is outside of the safety threshold. A service corresponding to the service data is rejected when the at least one type of the user's behavior data is outside of the safety threshold.