Edge-Enabled Lightweight Deep Learning Framework for Real-Time IoT Intrusion Detection

الملخص

The rapid expansion of Internet of Things (IoT) technologies has significantly increased cybersecurity risks associated with interconnected smart devices operating in resource-constrained environments. Conventional intrusion detection systems often fail to satisfy the real-time and low-resource requirements of modern IoT infrastructures due to high computational complexity, excessive memory usage, and cloud dependency. This research proposes a lightweight deep learning-based intrusion detection framework designed for real-time IoT network security with improved computational efficiency and edge deployment feasibility. The proposed system integrates efficient data preprocessing, lightweight feature engineering, and optimized deep learning classification to accurately distinguish between normal and malicious network traffic while minimizing resource consumption. Multiple publicly available cybersecurity datasets, including CICIDS2017, UNSW-NB15, ToN-IoT, and IoT-23, were utilized to evaluate the framework under diverse attack scenarios. Experimental analysis demonstrated that the proposed lightweight model achieves competitive intrusion detection accuracy while significantly reducing inference latency, memory usage, and computational overhead compared to conventional deep learning approaches. The edge-oriented architecture further improves real-time responsiveness, scalability, and privacy preservation by enabling localized traffic analysis without continuous cloud communication. The obtained results confirm that lightweight artificial intelligence techniques combined with edge computing technologies provide an effective and scalable solution for securing next-generation IoT environments against evolving cyber threats.