Research On Intrusion Detection in the Internet of Things Based on a Fusion Model

DOI: https://doi.org/10.62517/jes.202602109

Author(s)

Hu Shaojie, Liu Fengchun*, Han Yang

Affiliation(s)

School of Science, North China University of Science and Technology, Tangshan, Hebei, China *Corresponding Author

Abstract

Aiming at the high-dimensional and strongly time-dependent characteristics of network traffic data, this paper proposes a multimodal feature fusion model based on Temporal Convolutional Network (TCN) and Transformer architectures. An optimal 10-dimensional feature subset is constructed by integrating a Random Forest model with Recursive Feature Elimination (RFE). A dual-channel feature extraction architecture is designed: the TCN module captures local temporal patterns using dilated causal convolutions with residual connections, while the Transformer module models global dependencies through a self-attention mechanism. Furthermore, the model structure is optimized with residual connections to enhance information flow. the trade-off between complexity and efficiency is balanced by adjusting the TCN channel parameters ([64, 128]) and reducing the Transformer dimension (d_model = 8). Experimental results demonstrate that the proposed model achieves a detection accuracy of 98.7% on the UNSW-NB15 dataset, outperforming conventional single-model approaches by approximately 9.8%. This study provides a novel technical pathway for intrusion detection in complex network environments.

Keywords

Network Intrusion Detection; Feature Selection; Temporal Convolutional Network; Self-Attention Mechanism; Multimodal Fusion

References

[1] Sumedha Seniaray, and Rajni Jindal. "Performance Analysis of Anomaly-Based Network Intrusion Detection Using Feature Selection and Machine Learning Techniques. "Wireless Personal Communications prepublish(2024):1-31. doi:10.1007/S11277-024-11602-5. [2] Kim Czangyeob, et al. "Intrusion Detection Based on Sequential Information Preserving Log Embedding Methods and Anomaly Detection Algorithms. "IEEE ACCESS 9. (2021):58088-58101. doi:10.1109/ACCESS. 2021.3071763. [3] Maseno Elijah M., and Wang Zenghui. "Hybrid wrapper feature selection method based on genetic algorithm and extreme learning machine for intrusion detection. "Journal of Big Data 11.1(2024):doi:10.1186/S40537-024-00887-9. [4] Al-Yaseen Wathiq Laftah, Idrees Ali Kadhum, and Almasoudy Faezah Hamad. "Wrapper feature selection method based differential evolution and extreme learning machine for intrusion detection system. "Pattern Recognition 132. (2022):doi:10.1016/J. PATCOG. 2022.108912. [5] Johnson Singh Khundrakpam, Maisnam Debabrata, and Chanu Usham Sanjota. "Intrusion Detection System with SVM and Ensemble Learning Algorithms. "SN Computer Science 4.5(2023):doi:10.1007/S42979-023-01954-3. [6] RongPeng Yan. "NETWORK INTRUSION DETECTION BASED ON RANDOM FOREST ALGORITHM. "Journal of Computer Science and Electrical Engineering 7.3(2025):doi:10.61784/JCSEE3056. [7] Abdullah Mujawib Alashjaee. "Deep learning for network security: an Attention-CNN-LSTM model for accurate intrusion detection. "Scientific Reports 15.1(2025):21856-21856. doi:10.1038/S41598-025-07706-Y. [8] Hadiseh Rezaei, et al. "Federated RNN for Intrusion Detection System in IoT Environment Under Adversarial Attack. "Journal of Network and Systems Management 33.4(2025):82-82. doi:10.1007/S10922-025-09963-8. [9] Johnatan Alves de Oliveira, Anderson Fernandes Pereira dos Santos, and Ronaldo Moreira Salles. "RNN for intrusion detection in digital substations based on the IEC 61850. "Journal of Information Security and Applications 94. (2025):104197-104197. doi:10.1016/J. JISA. 2025.104197. [10] Kuldeep Singh. "Industrial internet of things fortify: multi-domain feature learning framework with deepdetectnet++ for improved intrusion detection. "Computers & Security 156. (2025):104506-104506. doi:10.1016/J. COSE. 2025.104506. [11] Muhammad Umer, et al. "Network intrusion detection model using wrapper based feature selection and multi head attention transformers. "Scientific Reports 15.1(2025):28718-28718. doi:10.1038/S41598-025-11348-5. [12] Yanchao Liu, et al. "TEGKT: tendency-enhanced evolution graph KAN transformer for information popularity prediction. "Journal of King Saud University Computer and Information Sciences 37.7(2025):206-206. doi:10.1007/S44443-025-00170-8. [13] Ibrahim A. Fares, et al. "TFKAN: Transformer based on Kolmogorov–Arnold Networks for Intrusion Detection in IoT environment. "Egyptian Informatics Journal 30. (2025):100666-100666. doi:10.1016/J. EIJ. 2025.100666. [14] Weidong Zeng, Hongshan Yang, and Yong Wang. "Intrusion Detection Based on Feature Selection and Transformer BiGRU". Abstracts of the 10th International Conference of Pioneering Computer Scientists, Engineers and Educators, ICPCSEE 2024 Part I. Ed. School of Computer Science and Technology, Shanghai University of Electric Power;, 2024, 149. doi:10.26914/c. cnkihy. 2024.079851.