Enhancing IoT RPL Protocol Security Against Black Hole Attacks with Deep Learning Techniques

Abstract

As the Internet of Things (IoT) extends its reach into critical infrastructures, it encounters escalating security threats, particularly black hole attacks that jeopardize network communication. The growing use of IoT in essential services underscores the need for robust network integrity. Driven by this urgency, our research has developed a sophisticated detection framework that effectively identifies black hole attacks within IoT networks utilizing the Routing Protocol for Low-Power and Lossy Networks (RPL). Our approach leverages the advanced pattern recognition capabilities of a deep Multi-Layer Perceptron (MLP) model, which has been rigorously trained and validated on a dataset generated under simulated conditions in Cooja. Key accomplishments of our study include achieving a high overall model accuracy of 94.3%, with specific accuracies of 94.4% for training, 94.2% for validation, and 94.0% for testing. The model exhibited minimal Mean Squared Error (MSE) values, with the lowest recorded validation MSE at approximately 0.029192. Additionally, the model's performance was marked by nearly perfect Receiver Operating Characteristic (ROC) curves, demonstrating Areas Under the Curve (AUC) close to 1 for both classes across all datasets. These performance metrics validate the model's efficacy in discerning the subtleties of black hole attacks, thereby enhancing network security analytics and contributing significantly to the proactive defense mechanisms against cyber threats in IoT networks. Our findings not only demonstrate the capabilities of deep learning models in cybersecurity but also underscore the importance of innovative solutions in safeguarding the expanding landscape of IoT infrastructures