next up previous
Next: Multiple Simultaneous Responses Up: Communication Paradigm Previous: Message Targeting/Distinguishing

Robustness to Active Interference

The only further difficulty related to an agent distinguishing which messages are intended for it arises in the presence of active interference. Consider a hostile agent H which hears a message that is directed to A at time t. H has full access to the message since all agents use the same communication channel. Thus if H remembers the message and sends an identical message at time u, agent A will mistakenly believe that the message is from a teammate. Although the message was appropriate at time t, it may be obsolete at time u and it could potentially confuse A as H intends.

This potential difficulty is avoided with the <encoded-time-stamp> field. Even a simple time stamp is likely to safeguard against interference since H is not privy to the locker-room agreement: it does not necessarily know which field is the time stamp. However, if H somehow discovers which field is the time stamp, it could alter the field based on the time elapsed between times t and u. Indeed, if there is a globally accessible clock, H would simply have to replace t with u in the message. However, the team can safeguard against such interference techniques by encoding the time-stamp using an injective function chosen as a part of the locker-room agreement. This function can use any of the other message fields as arguments in order to make decryption as difficult as possible. The only requirement is that a teammate receiving the message can invert the function to determine the time at which the message was sent. If the time at which it was sent is either too far in the past or in the future (according to the locker-room agreement), then the message can be safely ignored. Of course, it is theoretically possible for hostile agents to crack simple codes and alter the <encoded-time-stamp> field appropriately before sending a false message. However, the function can be made arbitrarily complex so that such a feat is intractable within the context of the domain. If secrecy is critical and computation unconstrained, a theoretically safe encryption scheme can be used. gif



next up previous
Next: Multiple Simultaneous Responses Up: Communication Paradigm Previous: Message Targeting/Distinguishing



Peter Stone
Mon Nov 24 11:31:14 EST 1997