Edge-AI Framework for Intrusion Detection in IIoT Networks Using Enhanced Deep Convolutional Neural Networks

Industrial Internet of Things (IIoT) networks are highly vulnerable to various cyber-attacks due to their interconnected nature and limited security mechanisms. This paper presents a novel Edge-AI-based Deep Convolutional Neural Network (EDCNN) framework for intrusion detection in IIoT networks. Our proposed system uses edge computing to process data close to its source, resulting in low latency responses and optimal resource efficiency. The EDCNN model, which includes an improved self-attention mechanism, detects complex intrusion patterns like advanced persistent threats (APTs), distributed denial-of-service (DDoS) attacks, and zero-day vulnerabilities. Additionally, we integrate Electronic Health Records datasets to validate the model’s adaptability across diverse IIoT applications, with experimental findings demonstrating the proposed framework’s exceptional accuracy of up to 98 percent, along with its scalability and computing economy, exceeding state-of-the-art techniques.