Blockchain-enabled depth-wise separable dual convolutional neural network for reliable scheduling to maximize network lifetime in wireless sensor network

Abstract Wireless Sensor Networks (WSNs) are a well-known and common deployed network in the field of monitoring and data collection in distributed systems, but the network has many difficulties including limited energy resources, reduced lifetime of the deployed network, inefficient scheduling, and unreliable communication. In this paper, the authors proposed a Blockchain based Dual Convolutional Neural Network (DS2CNN) approach to address the above issues, which is designed to effectively schedule tasks in WSNs and optimize the network lifetime. The proposed approach begins by a corona based zone partitioning model which organizes the network to enable its efficient management of data and reduces communication overhead. A Self-Guided Deep Clustering Network (SeGDCN) is proposed to learn discriminative representations and generate compact and balanced clusters out of the collected data. To perform energy-efficient and load-balanced operation of the network, cluster head (CH) selection using Improved Exponential Distribution Optimizer (I-EDO) is then carried out. The node activity scheduling is completed after clustering, using the DS2CNN model combined with a Dual Attention Transformer Network (DATN) to make intelligent judgments on the activity and sleep states under the condition of the network. To ensure security and trustworthiness, a Weighted Byzantine Fault Tolerance Consensus Algorithm (WByFTCA) is validated of the scheduling and control decisions by a mechanism based on blockchain, so that the integrity and bad behavior resistance is guaranteed. Finally, the optimal routing is done by implementing a Delay-sensitive Energy-Efficient Routing Scheme (DEERS), which uses residual energy, link quality and delay to select optimum routing. The experimental results demonstrate the effectiveness of the proposed framework over the conventional approaches which are potentially suitable for energy efficient and secure WSN application in terms of achieving an increase in network lifetime and decrease in energy consumption and latency.