Traditional security and access control systems, such as MLS/Bell-LaPadula, RBAC are rigid and do not contain automatic mechanisms through which a system can increase or decrease users' access to classified information. Therefore, in this paper, we propose a risk-based decision method for an access control system. Firstly, we dynamically calculate the trust and risk values for each subject-object pair. Both values are adaptive, reflecting the past behavior of the users with particular objects. The past behavior is evaluated based on the history of reward and penalty points. These are assigned by the system after the completion of every transaction. Secondly, based on the trust and risk values, an access decision is made.