Personal context information is important to enable context-aware applications in pervasive spaces. As privacy, many investigators have pointed out that it must be in the controllable state by using personal privacy policy. The long term objective of this research was to extract privacy policy from personal context information database with machine learning method. A context-aware telephone system was designed to collect data, learn policy and examine the accuracy of policy. In the step of PDA simulation, the results showed that the method was possible to make suitable decision for telephone communication in consideration of privacy protection. In order to further validate the proposed privacy protection method, the pervasive space was designed and implemented to get more real data about using context-aware telephone. At the step of pervasive space simulation, we design and code the prototype of context-aware telephone system from the view of software engineering.