A Deep Hybrid Learning Model For Intrusion Detection In IoT Network Traffic Using CNN- Transformer With PSO Tuning

Abstract— The rapid expansion of the Internet of Things
(IoT) has introduced significant security challenges due to the
increasing volume and diversity of network traffic generated by
interconnected devices. Traditional intrusion detection systems
often struggle to effectively capture complex spatial and
temporal patterns present in IoT network data. To address this
limitation, this paper proposes a deep hybrid learning
framework that integrates Convolutional Neural Networks
(CNN) and Transformer architectures for efficient intrusion
detection in IoT network environments. The CNN component is
employed to extract local spatial features from network traffic
data, while the Transformer module captures long-range
dependencies and sequential relationships to improve detection
accuracy. Furthermore, Particle Swarm Optimization (PSO) is
applied to optimize key hyperparameters of the hybrid model,
enhancing overall performance and stability. The proposed
model is evaluated using standard performance metrics
including accuracy, precision, recall, F1-score, and Receiver
Operating Characteristic–Area Under Curve (ROC–AUC).
Experimental results demonstrate that the proposed CNN–
Transformer model with PSO tuning achieves superior
detection capability, obtaining an AUC value close to 0.999 and
outperforming several conventional machine learning and deep
learning approaches. The results highlight the effectiveness of
the proposed hybrid framework in improving intrusion
detection performance and providing a reliable security solution
for modern IoT networks.