In order to restrict the behaviors of applications, a permission system is designed in Android system. However, for the permissions granted by the users, applications will no longer be restricted and can use these permissions at will, which may cause the potential permission abuse attacks. To detect the permission abuse behaviors of applications, an association analysis based detection method was proposed. This method dynamically detects sensitive behaviors of applications and operations of users, then calculates the degree of association between them. Detection result will be obtained through comparing the differences between detected applications and benign applications. A prototype system named DroidDect was designed and implemented based on the above method. The experimental results show that permission abuse behaviors in Android applications can be effectively detected by DroidDect with advantages including low system overhead.