Secure IoT Networks: A Deep Learning-Based Framework for Detecting Black Hole Attacks in RPL Protocol with Explainable AI and Edge Computing Integration

Abstract

With the rapid proliferation of the Internet of Things (IoT) across various sectors, including healthcare, smart cities, and critical infrastructure. IoT networks face heightened cyber threats, especially Black Hole Attacks (BHA). These attacks disrupt communication by maliciously routing data packets into Black Holes (BH) within the network, rendering critical information inaccessible and severely compromising network reliability and security. As IoT applications increasingly underpin essential services, establishing robust mechanisms to ensure network integrity becomes imperative. This study offers a sophisticated security framework to identify and prevent BHA in IoT networks that make use of the Routing Protocol for Low-Power and Lossy Networks (RPL) in order to address this challenge. Using the excellent accuracy of a Multi-Layer Perceptron (MLP) Neural Network (NN) model, this framework is a Deep Learning (DL)-based detection system. The Integration of Explainable Artificial Intelligence (AI) (I-XAI) approach is a noteworthy aspect of the methodology. By incorporating XAI, the framework achieves high detection accuracy and provides interpretable insights into the model’s Decision-Making (DM) process, addressing the often-cited black box issue in DL. The explainability of our model aids security analysts in understanding the specific patterns and characteristics that contribute to black hole attack detection, enhancing the reliability of our solution in real-world applications. To fortify the detection mechanism, we enhanced the framework with Real-Time (RT) anomaly detection capabilities enabled by advanced Edge-Computing (EC) devices. This allows rapid identification and response to suspicious activity, reducing latency and minimizing network vulnerability. Additionally, the framework incorporates Federated Learning (FL), enabling decentralized model updates across IoT nodes while preserving data privacy, an essential feature for compliance with emerging data protection regulations. A critical addition to our framework is a Trust Evaluation Mechanism (TEM), which assesses the trustworthiness of IoT nodes based on their behavior and historical data. This mechanism helps in dynamically adjusting the network trust levels and improves the accuracy of Attack Detection (AD) by correlating anomalous activities with trust scores. This multi-faceted approach ensures robust, transparent, and adaptive protection against BHA in IoT environments.