Defending Critical Infrastructure
Interfaces
Proceedings of the 7th international joint conference on Autonomous agents and multiagent systems: industrial track
| Hi-index | 0.00 |
The federal government's allocation of resources for defense against potential attacks generally involves depending on a multitiered organization consisting of federal, state, and local agencies for information on the risks faced in their jurisdictions and the costs and benefits of defensive actions they may take. These agencies then receive resources, ideally based on the collectively reported risks and resource needs. With private information about local risks and defensive actions and limited resources at the federal level, there are ample opportunities for agencies to take advantage of such a system for their benefit. This yields a suboptimal allocation of limited defensive resources. In this paper we describe this allocation problem formally as a game between a single principal (e.g., Congress) and agents representing more local agencies. This differs substantially from the treatment of the attacker--defender problem in the literature, where the defender is treated as a single, unified decision maker. We show that ignoring the within-team defender interactions in modeling the attacker--defender game leads to a suboptimal resource allocation. Existing results from agency theory are applied to this problem together with new results and insights into the interactions between an attacker and a multilevel defender.