Detection of Selfish Node in Mobile Ad Hoc Network by Adaptive Multi-Serial Cascaded Network

Abstract

Nodes are communicated not including the requirement of centralized organization or permanent transportation in mobile ad hoc networks (MANETs). The network topology frequently changes in the network poses several scalability challenges. Hence, an efficient selfish node recognition utilizing deep learning is implemented to overcome these issues. In the MANET topology, the network comprises multiple nodes that are responsible for routing, communication, and data transmission. The selfish nodes refuse to relay information to neighboring nodes. The occurrence of selfish nodes can significantly decrease system performance. This paper investigates certain input node attributes like hop count, residual energy, cooperation history, and co-operation rate, where the system is taken as the target co-operation rate. These considered node attributes are subjected to the adaptive multi-serial cascaded network (AMSCNet) for finding the selfish node present in the system; this network is composed of conditional autoencoder (CAE), deep temporal convolution network (DTCN), and deep capsule network (Deep CapsNet). To evaluate the model's effectiveness, the hyper-parameters in AMSCNet are optimized using hybridized Ebola and gold rush optimizer (HE-GRO) as Ebola optimization strategy (EOS) and gold rush optimizer (GRO). From the result analysis, the investigated HE-GRO-AMSCNet-based selfish node detection model achieved greater precision of 44.59% than CAE, 10.01% than DTCN, 31.27% than Deep_CapNet, and 9.12% than CAE_DTCN_Deep_CapNet. The efficacy of the offered self-node detection organization is compared with several existing systems.