On Mobile Agents Resistance to Traffic Analysis

This paper will concern itself with a formulation of a traffic analysis problem for mobile agents. It is an interesting theoretical problem as well as a critical feature when using agents on a massive scale in decision making systems. The decision making systems are applied to demanding and complex environments such as stock markets. The mobile agents used are natural targets for attacks because they provide information for decision making. The resulting information can have a value measured in millions of dollars and information of such a high value attracts potential attacks. An efficient way to attack the user of decision making systems is to learn her strategy and respond in kind. In this respect even passive observation of agents can provide useful data, namely what information they are gathering. A common first defense is to provide anonymity for mobile agents. However, what happens when anonymity is gone? What information then becomes available and what steps will the user take? Yet, the problem has not been previously formulated for such a framework. We formulate it in terms of various factors used for traffic analysis. These factors originate from different side channels that provide information on the operating agents. At the end we state a paradox, which links an excessive use of countermeasures against traffic analysis with weakening system security.