We consider the problem of achieving uniform and ergodic peer sampling in large scale open systems under adversarial behavior. The main challenge is to guarantee that any honest peer is able to construct a uniform and non-denitive (ergodic) sample of the peers identiers in the system, and this even in the presence of Byzantine peers controlled by the adversary. This sample is built out of a stream of peer identiers received at each node. We consider and study two types of adversary; an omniscient adversary that has the capacity to eavesdrop on all the messages that are exchanged within the system, and a blind adversary that can only observe messages that have been sent or received by peers he controls. In both models, the adversary can disrupt the input stream by injecting new messages or dropping messages sent by honest peers. Given any sampling strategy, we quantify the minimum eort an adversary has to exert on any input stream to prevent the sampling strategy from outputting a uniform and ergodic sample. We derive lower bounds for both adversary models.