A Hybrid CNN-LSTM-Transformer Model for IoT Networks Anomaly Detection

As Internet of Things (IoT) networks continue to expand in scale and complexity, there is an increasing need for autonomous anomaly detection (AD) systems that utilize highly accurate algorithms based on Machine Learning (ML) or Deep Learning (DL). The substantial volume of data generated by IoT devices makes DL particularly suitable for AD tasks, as it can autonomously determine significant features, identify meaningful data patterns, and detect anomalies in large datasets. This study proposes a hybrid deep learning model combining the capabilities of Convolutional Neural Networks (CNN) in capturing local spatial features, Long Short-Term Memory (LSTM) networks in modeling short-term temporal dependencies, and the Transformer's attention mechanism for enhanced contextual understanding and efficient parallel processing. Experiments using the IoT-DS-2 dataset demonstrate that the proposed CNN-LSTM-Transformer model significantly outperforms other hybrid architectures built from CNN, LSTM, and Transformer components, achieving superior classification accuracy, loss, precision, recall, and F1-Score.