Comparative analysis of deep learning and transformer-based approaches for HTTP web attack detection
Radiotekhnika · 30.09.2026 · (226) · 85-98
DOI: 10.30837/rt.2026.3.226.07 ↗Kurz gefasst
Fachartikel in der Zeitschrift Radiotekhnika, online erschienen am 30.09.2026. Verfasst von M.S. Kavetskiy und 1 weiteren Autor:innen.
Abstract (Originalsprache)
The work focuses on a comprehensive analysis and systematization of modern web attack detection methods in HTTP traffic using deep learning (DL) and Transformer-based natural language processing (NLP) techniques. The relevance of the review study is underscored by the urgent need to transition from traditional signature-based WAFs to intelligent defense systems. Unlike most existing works that evaluate artificial intelligence models solely by overall accuracy metrics, this article critically examines research solutions through the prism of their practical applicability for deployment in Production environments. The analyzed previous studies revealed a number of fundamental limitations in existing architectures, notably: unacceptable latency during real-time operation, the «black box» problem (lack of explainability of decisions), critical vulnerability to adversarial attacks, and a tendency to overfit on unbalanced datasets. In the context of these challenges, the review highlights the most mature engineering approaches, such as optimized cascaded architectures with lightweight pre-filtering modules and frameworks with automatic generation of interpretable defense rules. Based on the comparative analysis conducted, key vectors for future research are formulated, making this work a potentially important tool for developers of reliable, transparent, and optimized next-generation cybersecurity systems.
Zitieren
APA
Kavetskiy, M., & Ruzhentsev, V. (2026). Comparative analysis of deep learning and transformer-based approaches for HTTP web attack detection. Radiotekhnika, 85-98. https://doi.org/10.30837/rt.2026.3.226.07
Vancouver
Kavetskiy M, Ruzhentsev V. Comparative analysis of deep learning and transformer-based approaches for HTTP web attack detection. Radiotekhnika. 2026:85-98. doi:10.30837/rt.2026.3.226.07
BibTeX
@article{Kavetskiy2026,
author = {Kavetskiy, M.S. and Ruzhentsev, V.I.},
title = {Comparative analysis of deep learning and transformer-based approaches for HTTP web attack detection},
journal = {Radiotekhnika},
year = {2026},
number = {226},
pages = {85--98},
doi = {10.30837/rt.2026.3.226.07}
}
EUVIMED