Deep Q Siamese convolutional network for intrusion detection in WSN

Securing data transmission in Wireless Sensor Networks (WSNs) becomes challenging due to the presence of malicious nodes. As a solution, this framework introduces a Deep Q-based Siamese Convolutional Network (DQSC-Net) specifically created to identify intrusions in Wireless Sensor Networks. First, the WSN is simulated. Next, the Kookaburra Crayfish Optimization Algorithm (KCOA) is employed to select the Cluster Heads (CHs), where KCOA combines the Kookaburra Optimization Algorithm (KOA) and the Crayfish Optimization Algorithm (COA). Thus, the framework employs Blockchain (BC) for routing, and the Fractional Kookaburra Crayfish Optimization (FKCO), a hybrid of Fractional Calculus (FC) and KCOA is used to determine the next hop. Following this, the base station detects intrusions through the subsequent steps. The system begins by supplying the input log data to the data normalization module. To ensure consistency, a linear normalization method is employed at this stage to scale the data uniformly. Following this, information gain is utilized to select the most influential features. Ultimately, the proposed DQSC-Net, which integrates DQN with an SCNN, carries out the intrusion detection process. The framework delivers outstanding results, with a True Positive Rate (TPR) of 94.675%, True Negative Rate (TNR) of 90.766%, accuracy of 92.877%, False Alarm Rate (FAR) of 9.234%, False Negative Rate (FNR) of 5.325%, precision of 91.778%, F1-Score of 93.204%, and balanced accuracy of 92.721%.