A reference monitor is a specific set of requirements governing the reference validation mechanism in a computer's operating system code. It acts to provide a theoretical ideal which protects the computer against potential hackers who might attempt to interfere with its access control policy. If the code of an operating system fails to meet the standards set by its reference monitor, the system will be vulnerable to external threats from malicious users seeking to breach its security policies.
A computer's access control policy determines not only who has the authority to log in to the system, but the individual users' rights to modify elements of the computer's system. It establishes a working hierarchy between authorized administrators of the computer — who can add and remove programs, make significant changes to the operating system, and so on — and regular restricted users of the computer, who are likely to have less freedom to modify the computer's systems. The reference validation mechanism sets these policies; in effect, it is like a security officer at a club or business. It does the dirty work of ensuring that everyone who gets into the system has the right to be there, along with keeping everyone in their place.
If the reference validation mechanism is a security officer, the reference monitor is the executive who determines the most effective paradigm for security. The reference monitor has three overarching goals for the reference validation mechanism. These can be generally summarized as alertness, incorruptibility, and testability.
Alertness means that the programmers of the operating system must ensure that the reference validation mechanism is active at all times. In other words, it has to remain "on the job" 24 hours a day in order to remain effective. If the operating system is running, the reference validation mechanism must be awake and working.
Incorruptibility means that the reference control mechanism must remain insusceptible to tampering. This means that it must ignore any and all outside attempts to influence its behavior. The goal of this is to prevent hackers from compromising the security of the system.
Finally, testability means that it must be a small enough process that the operating system can monitor it at all times. According to the ideal reference monitor policies, the operating system should be able to test the reference validation mechanism freely, checking to see whether it is doing its job correctly. These tests must be verifiable by the operating system, allowing it to run the tests repeatedly and achieve consistent, accurate results.