Quality of Service (QoS) of disadvantaged networks is considered from a purely network standpoint in existing works. Adversarial intervention in such networks is not analyzed, nor is it possible to infer if a QoS loss is benign or otherwise. In this paper, we present a QoS loss inference module, where the end nodes can infer the nature of a QoS loss in a non-intrusive manner. The objective of this work is to develop a conceptual framework to model the inference module, and investigate its integration into existing platforms. We abstract the problem of link selection (as opposed to route selection) in disadvantaged networks as a resource selection problem, and apply a game theoretic model to set limits on the rate of convergence. Using this convergence rate, our loss inference model can distinguish between adversarial network manipulation and benign network loss. Such a module will help manage the operation of disadvantaged networks in a more effective manner in critical networks.