In this paper, we use reliability theory to determine the answer of an important question about how often we need to update a piece of secure information. For example, one may need to determine an appropriate time for updating a new password. To generalize, we compute an efficient updating time for a dependable secure computing (DSC) system which stores secret information in a reliable and secure fashion among n servers. The secret information can be reconstructed by successfully connecting to any k servers (k ≤ n), but cannot be revealed by using (k − 1) or fewer servers. Generally, the secret information has to be updated in the system in order to prevent adversaries from learning secret information by breaking into k servers in a DSC system. There are serval proposed procedures to update secret information, but few has been able to determine what an appropriate updating time on a DSC system is. To achieve the most reliable way to safeguard the secret information, we developed an interesting approach to compute the updating time on a DSC system from the reliability point of view.