ARA-MAC: Attacker identification using logistic regression

Selfish behavior by any node/terminal in a network at MAC layer can have damaging effects on the overall performance of the wireless networks. In IEEE 802.11, the distributed coordination function protocol is Carrier-sensing multiple-access with collision avoidance (CSMA/CA)-based. Due to its ease of implementation it is widespread. Since the terminals/nodes accessing these networks are independent operators, they might not abide by the rules and will try to gain an unfair access to the network for selfish reason or just to decrease the network performance. So, this paper improves upon an already existing MAC layer Protocol known as ARA-MAC protocol, which gives improved performance as compared to CSMA/CA protocol in the presence of attackers. For this we implement a machine learning technique called logistic regression for the classification of nodes into attacker and non-attacker. Using this we improve upon the previous attacker detection Algorithm used in ARA-MAC protocol for ameliorating the Quality of Service (QoS) for each and every node in the network.