Since all CRISIS certificates contain timeouts and since these certificates are distributed across the wide area, the system must make assumptions about clock synchronization. Further, CRISIS must protect against security attacks exploiting a node's notion of time. If time on a machine is corrupted, statements can be used beyond their period of validity.
Today, most workstations possess fairly accurate clocks that are periodically synchronized with any of a number of external sources. However, time-sensitive applications (and hence, reference monitors) may require guarantees above and beyond such loose synchronization, for example, that the local clock is periodically synchronized with a trusted external source. Other applications will require an invoice of all assumptions made during a computation in the case where data is corrupted or leaked to determine the exact cause of the corruption, assuring complete accountability.
For CRISIS, we assume the presence of replicated, trusted time servers. Principals producing certificates with timeouts (e.g., CA's and OLA's) contact these servers periodically to obtain signed certificates containing the current time to validate the principal's notion of time. If the principal's time differs by more than a few seconds (i.e., within network delay bounds) from the time supplied by the server, the principal assumes that either the time server or the local operating system/hardware has been compromised (to determine which, a second server might be contacted). Such communication with time servers need not be synchronous, since the time certificates can be cached to prove recent synchronization.
In CRISIS, time certificates are provided to resource managers to prove that a node's notion of time closely matches the value reported by a trusted time server at some recent point in the past. CRISIS identity and transfer certificates report time values (such as expiration time) relative to the value contained in a chained time certificate. While use of time certificates does not guarantee that time-based attacks can be avoided or prevented, it can aid in determining the cause of certain security violations post-mortem. Thus, if a security breach is detected, analysis of certificates used to gain unauthorized access can be used to determine the cause of the attack. For example, examination of the certificates may show that a node attempted to use an expired time certificate or that a time server was compromised and reported faulty values of time.