An Optimized Deep Learning Based Malicious Nodes Detection in Intelligent Sensor-Based Systems Using Blockchain

Abstract

—In this research work, a blockchain-based secure routing model is proposed for Internet of Sensor Things (IoST), with the assistance acquired from deep learning-based hybrid meta-heuristic optimization model. The proposed model includes three major phases: (a) optimal cluster head selection, (b) lightweight blockchain-based registration and authentication mechanism, (c) optimized deep learning based malicious node identification and (d) optimal path identification. Initially, the network is constructed with N number of nodes. Among those nodes certain count of nodes is selected as optimal cluster head based on the two-fold objectives (energy consumption and delay) based hybrid optimization model. The proposed Chimp social incentive-based Mutated Poor Rich Optimization (CMPRO) Algorithm is the conceptual amalgamation of the standard Chimp Optimization Algorithm (ChOA) and Poor and Rich Optimization (PRO) approach. Moreover, blockchain is deployed on the optimal CHs and base station because they have sufficient storage and computational resources. Subsequently, a lightweight blockchain-based registration and authentication mechanism is undergone. After the authentication of the network, the presence of malicious nodes in the network is detected using the new Optimized Deep Belief Network. To enhance the detection accuracy of the model, the hidden layers of Deep Belief Network (DBN) is optimized using the new hybrid optimization model (CMPRO). After the detection of malicious nodes, the source node selects the shortest path to the destination and performs secure routing in the absence of malicious node. In the proposed model, the optimal path for routing the data is identified using the Dijkstra algorithm. As a whole the network becomes secured. Finally, the performance of the model is validated to manifest its efficiency over the existing models