In this paper, we introduce the decentralized fault detection filter, a structure that results from merging decentralized estimation theory with the game theoretic fault detection filter. A decentralized approach may be the ideal way to health monitor large-scale systems, since it decomposes the problem down into (potentially smaller) “local” problems. These local results are then blended into a “global” result that describes the health of the entire system. The benefits of such an approach include added fault tolerance and easy scalability. An example given at the end of the paper demonstrates the use of this filter for a platoon of cars proposed for advanced vehicle control systems.