The main objective of this paper is to design Evolutionary Game Theory (EGT) based overlay network structure with the combination of network reciprocity and group selection mechanisms to achieve evolutionary cooperation among individual nodes. Each node in the proposed network structure has ability to detect the DDoS attacks using Non-parametric sliding window based change point detection algorithm. It analyzes every incoming traffic behavior changes and generates the attack patterns based on the traffic profile. The proposed network structure is considered as a tree like structure in the experimental analysis, which monitors the network traffic and provides faster detection during the distributed flooding attack. As a result, our network detects the DDoS attacks at the intermediate stage itself with the sharing nature of attack patterns, without any interruption to legitimate users. Moreover, it provides excellent detection accuracy (98%) and high degree of efficiency during high probability of network traffic. © 2012 IEEE.