Deep Learning‐Based Intrusion Detection in Encrypted Network Traffic: A Systematic Review of Architectures, Datasets and Research Gaps
IET Networks · 2026-01 · 15 (1)
DOI: 10.1049/ntw2.70039 ↗Kurz gefasst
Systematische Übersicht in der Zeitschrift IET Networks, online erschienen am 05.11.2026. Verfasst von Dadzie Ekow Abraham und 3 weiteren Autor:innen. Die Arbeit nennt 76 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The increasing adoption of encryption protocols such as Transport Layer Security (TLS) and QUIC has strengthened data privacy but reduced the effectiveness of traditional intrusion detection systems (IDS) that rely on payload inspection. Deep learning offers an alternative by learning discriminative patterns from encrypted traffic without requiring decryption. This study systematically reviewed deep learning‐based intrusion detection in encrypted network traffic following the PRISMA 2020 framework. From 980 identified records, 50 peer‐reviewed studies published between January 2020 and December 2025 were included and systematically analysed. The findings indicated strong overall detection performance, with 45 of the 50 studies (90%) reporting a representative accuracy or F1‐score of at least 90%. Hybrid architectures emerged as the dominant model category, representing 58% of the reviewed studies, reflecting an increasing tendency to combine complementary spatial, temporal, attention, and other representation‐learning mechanisms. Despite these advances, persistent challenges included reliance on limited benchmark datasets, class imbalance, weak cross‐dataset generalisation, inconsistent evaluation practices, computational complexity, and limited real‐time deployment and explainability. To consolidate the evidence and guide future research, the review proposes a six‐dimensional taxonomy and an end‐to‐end conceptual framework for encrypted traffic IDS. Overall, the findings highlight the need to move beyond benchmark accuracy towards generalisable, computationally efficient, robust, explainable, and deployment‐ready intrusion detection systems for real‐world encrypted network environments.
Zitieren
APA
Abraham, D. E., Kwabena, O., Eric, O. O., & Linda, B. A. (2026). Deep Learning‐Based Intrusion Detection in Encrypted Network Traffic: A Systematic Review of Architectures, Datasets and Research Gaps. IET Networks, 15(1). https://doi.org/10.1049/ntw2.70039
Vancouver
Abraham DE, Kwabena O, Eric OO, Linda BA. Deep Learning‐Based Intrusion Detection in Encrypted Network Traffic: A Systematic Review of Architectures, Datasets and Research Gaps. IET Networks. 2026;15(1). doi:10.1049/ntw2.70039
BibTeX
@article{Abraham2026,
author = {Abraham, Dadzie Ekow and Kwabena, Owusu‐Agyemang and Eric, Osei Opoku and Linda, Banning Amoako},
title = {Deep Learning‐Based Intrusion Detection in Encrypted Network Traffic: A Systematic Review of Architectures, Datasets and Research Gaps},
journal = {IET Networks},
year = {2026},
volume = {15},
number = {1},
doi = {10.1049/ntw2.70039}
}
EUVIMED