Spyware is considered a great threat to confidentiality since it can cause loss of control over private data for computer users. This kind of threat might select some data and send it to another third party without the consent of the user. Spyware detection has been presented traditionally by three approaches signature based detection, behavior based detection and specification based detection. These approaches were successful in detecting known spyware but have proven failure in detecting unknown spyware. In this paper we introduce a framework for anti spyware system that combine data mining and design patterns in detecting and classifying spyware. The proposed anti spyware system can be reusable as it can modify itself for any new or modified spyware.