LMCA: a lightweight anomaly network traffic detection model integrating adjusted mobilenet and coordinate attention mechanism for IoT

Dezhi Han (), HongXu Zhou (), Tien-Hsiung Weng (), Zhongdai Wu (), Bing Han (), Kuan-Ching Li () and Al-Sakib Khan Pathan ()
Additional contact information
Dezhi Han: Shanghai Maritime University
HongXu Zhou: Shanghai Maritime University
Tien-Hsiung Weng: Providence University
Zhongdai Wu: Shanghai Ship and Shipping Research Institute Co., Ltd.
Bing Han: Shanghai Ship and Shipping Research Institute Co., Ltd.
Kuan-Ching Li: Providence University
Al-Sakib Khan Pathan: United International University (UIU)

Telecommunication Systems: Modelling, Analysis, Design and Management, 2023, vol. 84, issue 4, No 8, 549-564

Abstract: Abstract As widely known, most of the Internet of Things (IoT) devices own small storage and constrained computing power, and hence, their poor security evaluation capabilities make them vulnerable to several types of network attacks. Given this setting, anomaly network traffic detection techniques based on deep learning (DL) offer some practical solutions, and they have brought new opportunities to the security of the IoT. However, existing DL models for anomaly network traffic detection need better flexibility and classification accuracy. Also, the scale of those models needs to be optimized, as a sheer majority of them need to be more suitable for deployment on terminal devices of IoT. Therefore, we propose an anomaly network traffic detection model in this work LMCA, standing for Lightweight Model Integrating adjusted MobileNet and Coordinate Attention mechanism. Combining the adjusted MobileNet model and the coordinate attention mechanism, it constructs a lightweight anomaly network traffic detection model and effectively extracts traffic data's local, global, and spatial–temporal features, which would be easy to deploy on IoT terminals. LMCA has a small scale and good performance, making it suitable for IoT environments. Moreover, we use an original traffic feature extraction method to reduce redundant features and speed up neural network convergence. This work also solves a problem so that the original MobileNet model could perform better on a small dataset, extending the anomaly traffic detection for IoT. To simulate the IoT environment, we used the wired network dataset CICDS2017 and the wireless network dataset AWID. Experimental results demonstrate that the proposed work outperforms other existing methods, the accuracy reached 99.96% on the CICIDS2017 dataset and 99.98% on the AWID dataset.

Keywords: Anomaly network traffic detection; Coordinate attention mechanism; Deep learning; IoT; MobileNet model (search for similar items in EconPapers)
Date: 2023
References: View complete reference list from CitEc
Citations: View citations in EconPapers (1)

Downloads: (external link)
http://link.springer.com/10.1007/s11235-023-01059-5 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:telsys:v:84:y:2023:i:4:d:10.1007_s11235-023-01059-5

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/11235

DOI: 10.1007/s11235-023-01059-5

Access Statistics for this article

Telecommunication Systems: Modelling, Analysis, Design and Management is currently edited by Muhammad Khan

More articles in Telecommunication Systems: Modelling, Analysis, Design and Management from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().