On replay detection in distributed systems

Various approaches to the problem of replay detection in distributed systems are briefly reviewed. An approach based on combining a variable-size time-window mechanism with a challenge mechanism is proposed. This approach has the following properties: (1) it does not depend on clock synchronization, (2) it allows the setting of a minimum server's memory-buffer size in a way that ensures acceptance of all legitimate client requests and (3) it is robust without requiring stable (nonvolatile) memory for the server buffer needed to save past client requests.<
>