Semantic Evolution and Consistency Learning for Robust Malicious Network Traffic Detection
Engineering Reports · 2026-06 · 8 (6)
DOI: 10.1002/eng2.70862 ↗At a glance
Journal article in Engineering Reports, published online on 2026-06-02. Written by Jing Yang and 1 co-authors. 38 references. The full text is freely available (open access, CC BY 4.0).
Abstract (original language)
As network environments become increasingly complex, malicious traffic exhibits characteristics such as stealthy behaviors, multi‐stage attack processes, and dynamic semantics, posing significant challenges to existing detection methods. Most existing approaches treat malicious traffic detection as a static classification problem, which limits their ability to model semantic evolution and generalize under distribution shifts. To address this issue, this paper proposes the semantic evolution and consistency network (SECN), a robust malicious traffic recognition framework based on dynamic semantic modeling and structured constraints. SECN adopts a dual‐level temporal representation mechanism, in which micro‐level encoding captures short‐term behavioral patterns, while macro‐level modeling characterizes semantic evolution across attack‐stages. In addition, inspired by metric‐learning principles, this paper introduces a task‐adapted semantic consistency constraint to reduce intra‐class semantic dispersion and maintain inter‐class margins in the latent evolution space. Experimental results on the UNSW‐NB15 and IoT‐23 datasets show that, in terms of Accuracy, Precision, Recall, F 1‐score, and false positive rate (FPR), SECN improves detection robustness under class imbalance, unknown‐attack settings, and cross‐dataset distribution shifts.
Cite
APA
Yang, J., & Tan, W. (2026). Semantic Evolution and Consistency Learning for Robust Malicious Network Traffic Detection. Engineering Reports, 8(6). https://doi.org/10.1002/eng2.70862
Vancouver
Yang J, Tan W. Semantic Evolution and Consistency Learning for Robust Malicious Network Traffic Detection. Engineering Reports. 2026;8(6). doi:10.1002/eng2.70862
BibTeX
@article{Yang2026,
author = {Yang, Jing and Tan, Wei},
title = {Semantic Evolution and Consistency Learning for Robust Malicious Network Traffic Detection},
journal = {Engineering Reports},
year = {2026},
volume = {8},
number = {6},
doi = {10.1002/eng2.70862}
}
EUVIMED