Performance of Packet Loss Differentiation Model by using Cuckoo Search Back-Propagation Neural Network (CSBPNN) for Mobile Ad-hoc Networks

Abstract

Mobile ad-hoc networks (MANETs) comprise a compilation of mobile nodes which are linked through wireless connections in an infrastructure-less network. Many of the available trust-based security schemes for MANETs believe packet losses as an indicator of probable attacks by malicious nodes. Packet loss occurs due to different reasons namely interference, queue overflow, and node mobility. The process of detecting the actual reason for the occurrence of packet loss is an important security solution. This study develops a new Cuckoo Search Back-Propagation neural network (CS-BPNN) based packet loss differentiation mechanism. Since the classical BPNN suffers from the issue of local optima, cuckoo search (CS) algorithm is incorporated in the training stage of BPNN the convergence rate and avoids local optima problem. The presented CS-BPNN model is simulated and results demonstrate that the presented model enhances the packet loss discrimination as well as throughput with a reduction in congestion.