Access control is an important issue in information security. It is a necessary mechanism for protecting data in a computer system. In this paper, we apply support vector machines to construct an access control system. The security management framework is a pattern classification system using well trained support vector machines. The proposed system can not only be applied to broad access policies, but the access rights of users can also be associated with a time period. The proposed scheme is useful to make predictions on real-time data and it is better adapted to complex computer systems.