Deep learning-based intrusion detection for IoT networks: a scalable and efficient approach

The rapid expansion of the Internet of Things (IoT) has revolutionized industries by enabling seamless connectivity, but it has also introduced significant security vulnerabilities, making IoT networks prime targets for cyberattacks. Traditional intrusion detection systems often struggle to cope with the high volume and dynamic nature of IoT traffic, necessitating the development of more robust and intelligent security mechanisms. This research presents a deep learning-based approach for real-time threat detection in IoT networks, leveraging advanced models such as 1D convolutional neural networks (CNNs), long short-term memory (LSTM) networks, recurrent neural networks (RNNs), and multi-layer perceptrons (MLPs) to enhance intrusion detection. The study utilizes the CIC IoT-DIAD 2024 dataset, a comprehensive collection of flow-based network traffic containing both benign and attack scenarios. The proposed models were trained and evaluated on flow-based feature sets, optimizing hyperparameters to maximize accuracy, recall, and F1-score. In multi-class classification, 1D CNN achieved the highest accuracy of 99.12%, followed by LSTM (98.98%), RNN (98.43%), and MLP (97.21%). For binary anomaly detection, 1D CNN again demonstrated superior performance with an accuracy of 99.53%, while LSTM, RNN, and MLP achieved 99.52%, 99.25%, and 98.78%, respectively. The results indicate that 1D CNN is the most effective model for real-time IoT intrusion detection, excelling in feature extraction and attack classification. The findings contribute to the development of scalable and efficient deep learning-based security solutions, improving the ability to detect and mitigate cyber threats in IoT environments.