A secure routing protocol using hybrid deep regression based trust evaluation and clustering for mobile ad-hoc network

Mobile ad hoc networks (MANETs) are autonomous networks where mobile nodes are connected with one another over wireless links using multi-hop transmission. Malicious attacks can severely disrupt routing setup and data transmission, seriously compromising the security and normal functioning of MANETs. This paper introduces regression and clustering techniques to compute the trust values of nodes and improve the routing of transferred packets in the MANET. Initially, the trust value of each node is detected with the help of the routing paths’ reputation. However, the trust evaluation using mathematical methods is challenging due to the ambiguity of the attack and the network’s diversity. To tackle this issue, the proposed model transforms the computation of the node’s trust values into a linear regression problem and solves this problem using Stacked Autoencoder with Sparse Bayesian Regression (SAE-SBR). After evaluating the trust value, a private collaborative K-means (PCK-means) clustering approach is proposed for clustering the nodes into benign or malicious based on the obtained trust values of nodes. Finally, the path that contains few nodes in the low trust value group and medium trust value group is chosen as the optimal routing path. The simulation results show that the proposed method minimizes the end-to-end delay time to 0.023 s in the presence of 10 malicious nodes. Also, it achieved a maximum throughput of 95.568 KbpS, a maximum detection accuracy of 94% and a maximum packet delivery ratio of 99.5%.