Improvement and Verification of Deep Learning-Driven Network Anomaly Traffic Detection Model
Zhao Liangpan12  Wei Chaoxin12

1 Mongolian National University，Ulaanbaatar city, Bayangol District, 11th Khoroo, National University of Mongolia，16060；

2 Jiangxi Institute of Fashion Technology，No. 103, Lihu Middle Avenue, Xiangtang Development Zone, Nanchang City, Jiangxi Province，330201；

Abstract：In the context of digital transformation, network security faces severe challenges. Traditional anomaly traffic detection methods have significant limitations. Although deep learning offers a new direction for this field, existing models suffer from issues such as single-feature capture, weak scenario adaptability, and insufficient real-time response. This paper constructs an optimization system from three dimensions: "Feature Fusion - Architecture Design - Training Mechanism." Multi-dimensional feature fusion is achieved through "Micro - Meso - Macro" three-level extraction and dynamic weighting, addressing the problem of incomplete feature coverage. The LFA lightweight architecture is designed, simplifying modules and optimizing the preprocessing pipeline to enhance real-time response capability. Multi-scenario fusion training and dynamic adjustment strategies are adopted to improve the model's scenario adaptability and robustness. Multi-scenario verification shows that the optimized LFA model outperforms traditional deep learning models in detection accuracy, scenario adaptability, and real-time responsiveness. Finally, the study points out limitations regarding dataset coverage and the identification of sparse anomaly traffic, and suggests future research directions.

Keywords：Deep Learning; Network Anomaly Traffic Detection; Feature Fusion; Scenario Adaptability

References

[1] Yang Yuelin, Bi Zongze. Deep Learning Based Network Traffic Anomaly Detection [J]. Computer Science, 2021(S2): 540-546.

[2] Li Yanmiao. Research on Network Anomaly Traffic Detection Technology Based on Deep Learning [D]. Beijing University of Posts and Telecommunications, 2023.

[3] Hang Mengxin, Chen Wei, Zhang Renjie. Anomaly Traffic Detection Based on Improved 1D-CNN [J]. Journal of Computer Applications, 2021, 41(2): 433-440. DOI:10.11772/j.issn.1001-9081.2020050734.

[4] Yin Chuanlong. Research on Network Anomaly Detection Technology Based on Deep Learning [D]. Strategic Support Force Information Engineering University, 2019.

[5] Gao Jia. Research and System Implementation of Network Anomaly Traffic Detection Based on Deep Learning [D]. Shihezi University, 2023.