Squeezing State Spaces of (Attack-Defence) Trees

In earlier work, we presented translations of attack-defence trees (ADTrees) to extended asynchronous multi-agent systems. By avoiding some sequences, agent models constructed via these transformations already embed state space reductions. Here, we introduce Guarded Update Systems and their synchronisation topology, allowing us to define a new general reduction scheme that applies to tree topologies, and in particular to ADTrees. The reduction exploits the layered structure of a tree by avoiding unnecessary interleavings between nodes at different depths. We prove the soundness of this new method and present extensive experimental results, including scalable models, to demonstrate it can be effectively used alongside previously employed techniques.