Abstract
The testing of complex adaptive systems for emergent misbehaviours currently lacks any kind of automated support. In this thesis, we present Incremental Adaptive Corrective Learning, an evolutionary learning technique which automatically identifies vulnerabilities and performance-limiting behaviours induced by adversarially controlled agents inserted into a tested mobile ad-hoc network. Three case studies are presented, each demonstrating how Incremental Adaptive Corrective Learning is used to induce poor network performance in different mobile ad-hoc network applications.
