Journal article Open Access
Antonopoulos, Angelos; Verikoukis, Christos
The altruistic user behavior and the isolation of malicious users are fundamental requirements for the proper operation of any cooperative network. However, the widespread use of new communication techniques that improve the cooperative performance (e.g., network coding) hinders the application of traditional schemes on malicious users detection, which are mainly based on packet overhearing. In this paper, we introduce a cooperative nonparametric statistical framework, namely COPS, for the mitigation of user misbehavior in network coding scenarios. Given that the behavior of adversaries cannot be characterized by certain probability distributions, the proposed scheme exploits two well-known nonparametric statistical methods, i.e., Kruskal-Wallis analysis and Conover-Iman multiple comparisons, for the detection and identification, respectively, of malicious users in the network. It is worth noting that the COPS framework does not require monitoring of the wireless channel and additional overhead, as its operation is based on the processing of the existing control packets. We assess the performance of the proposed scheme in various scenarios, showing that COPS is able to effectively handle attacks in the network, even when malicious users adopt a smart probabilistic misbehavior.