Enhancing IoT Security Through a Hybrid Deep Learning Model: CNN Meets Transformer for Robust Intrusion Detection

The Internet of Things (IoT) is transforming sectors like healthcare and industrial automation, but its widespread deployment and resource-constrained devices make it vulnerable to cyberattacks. Conventional Intrusion Detection Systems (IDS) struggle to detect the evolving and complex attack strategies that exploit this vulnerability. We propose a hybrid deep learning architecture that integrates 1D Convolutional Neural Networks (CNNs) and transformer-based models to capture both spatial and temporal patterns in network traffic. The proposed method is trained and evaluated on the CIC-IDS2017 dataset, and we design a robust preprocessing pipeline including feature normalization, class balancing using SMOTE, and data cleaning to mitigate inconsistencies and class imbalance. In comparison to baseline models such as CNN-only, transformer-only, and LSTM-based systems, our hybrid model consistently outperforms them in terms of detection accuracy and recall, especially for minority-class attacks. Our results demonstrate a detection accuracy of 92.5%, suggesting that hybrid deep learning models can offer more reliable IDS performance in real-time IoT environments.