A Convolutional Neural Network (CNN)-Based Anomaly Detection Framework for Internet of Things (IoT) Systems Reinforced by Blockchain for Privacy and Safety

Abstract

A developing approach for identifying anomalies, especially in contexts like Internet of Things (IoT) systems, is the integration of Convolutional Neural Networks (CNN) with the Whale Optimization Algorithm (WOA). If you want to modify the CNN's parameters for better anomaly identification, you may apply the WOA, a strategy for optimization inspired by nature, in conjunction with the CNN's effectiveness in detecting spatial trends. Robust anomaly detection is critical for maintaining the authenticity, safety, and dependability of IoT systems in an ever-changing environment. We need more sophisticated and adaptable detection algorithms to handle the massive amounts and diverse types of data generated by the IoT. This research proposes a new approach to detect anomalies by integrating CNNs with the WOA, which should help with these issues. CNNs are a basic model for anomaly detection because of their famed capacity to extract spatial trends from complicated information. Learning rates, kernel sizes, and the number of layers are hyperparameters that have historically needed to be fine-tuned by hand in order to achieve optimal CNN efficiency. To tune these additional parameters successfully, we use the WOA, a metaheuristic influenced by nature and based on the bubble-net hunting tactic of humpback whales. The CNN's accuracy and resilience in recognizing anomalies tasks are improved by WOA's global search features, which allow it to obtain optimum configurations. To ensure the suggested technique worked, we ran comprehensive tests using real-world IoT datasets. The findings show that the CNN-WOA hybrid model outperforms traditional approaches in terms of speed of computation, recall, and accurateness, and it frequently maintains an accuracy of over 95% in identifying abnormalities. This precision demonstrates how well the model deals with unbalanced, noisy, and high-dimensional IoT data. This work offers an expandable approach for smart IoT system management by fusing deep learning with bio-inspired minimization. It establishes a precedent for future research in anomaly identification.