Abstract-In this paper we investigate the behavior of degreebased clustering algorithms with respect to their stability and attack-resistance. Our attack scenario tries to bias the clustering head selection procedure by sending faulty degree claims. We propose a randomized variant of the highest degree algorithm which is proved, through experimental results, attack-resistant without imposing significant overhead to the clustering performance. In addition, we extend our proposal with a cooperative consistent clustering algorithm which integrates security into the clustering decision achieving attacker identification and classification.

Abstract-In this paper we revisit some considerations relative to the performance of re-clustering algorithms in MANET. We recommend that for a secure MANET the design of re-clustering algorithms should not only provide for clustering stability, but also for network robustness in terms of network connectivity, of message reliability, of tolerance against the attacks that target the cluster head nodes and of tolerance to random node failures due to energy drains. We also take into account the possibility of malicious users that might thwart the network protocol by advertising false topology information. We propose a distributed mechanism that for unbiased cluster head election first demands certain levels of consistency to be reached among the nodes.

No abstract

No abstract

Abstract-In this study we evaluate with experiments three generic clustering algorithms, namely the Lowest-ID, the Highest Degree and the Extended Robust Re-clustering Algorithm which is the one proposed. The aim is to investigate which are the factors that have significant effect on the re-clustering performance. We isolate those performance factors as being network conditions that we simulate with a particular focus on the node deployment pattern, the mobility pattern, the radio transmission range and the energy of the ad hoc nodes. For the evaluation of the re-clustering efficiency and for the comparison of the three algorithms we examined conventional re-clustering performance metrics, such as the cluster head modification rate and the number of the generated clusters but also reliability metrics, such as the cluster head availability probability and the end to end message delivery ratio. We draw generic outcomes that hold for the three algorithms and we also discuss the behavior of the proposed algorithm.