To tolerate against Byzantine failures, a traditional system requires n = 3f + 1, where n is the number of nodes and f is the number of Byzantine nodes. More recently, a more efficient methodology has been proposed, requiring only n = 2f + 1 by separating agreement from execution. We argue that since nodes are prone to churn and the network topology is not often fully connected in a dynamic system (e.g., cloud computing), an unstructured form of achieving Byzantine fault tolerance is necessary such as using gossip. In such a case, however, due to the presence of malicious Byzantine nodes, existing membership management mechanisms are not suitable for preserving uniformity of random sampling. In this paper, we propose a novel unstructured membership management implementation for Byzantine fault tolerance in clouds. Analytic studies reveal that our implementation of membership management leads to a more adequate and satisfactory solution than existing ones.