EUVIMEDEUVIMED entdecken
Crossref

KAN-Vis: Efficient and Lightweight Visual Technique for Network Traffic Classification Using Kolmogorov–Arnold Network

Anurag Dutta, Pallavi Anand, Sangita Roy, Rajat Subhra Chakraborty

IEEE Transactions on Networking · 2026 · 34 · 6595-6605

DOI: 10.1109/ton.2026.3722338 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Networking, 2026. Verfasst von Anurag Dutta und 3 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Gefördert von: Defense Research and Development Organization (DRDO), Ministry of Defense, Government of India.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Network traffic classification is a problem of fundamental importance in network security, and has been widely researched. However, the complexity of the problem has increased substantially in recent years, since most modern network traffic is encrypted end-to-end, and often further protected using VPN , and anonymization schemes like Tor . Existing methods have extracted important statistical features from the network traffic flow and used them to classify the traffic using machine learning. In this research, a set of four visual network traffic classifiers ( KAN-Vis ) based on the novel Kolmogorov-Arnold Network architecture is proposed. The input to these models will be a pictorial representation of the time series of packet-size histograms. The models were evaluated across five benchmark datasets. On ISCX datasets, KAN-Vis outperforms state-of-the-art techniques by 2.03%, 0.52%, and 1.98% for regular, Tor -encrypted, and VPN -protected traffic flows, respectively. KAN-Vis models also generalize for modern encryption protocols, where they achieve a 3.24% improvement over the best of existing ensemble models on CIC-IDS-2018 botnet detection, a 0.13% improvement over the state-of-the-art CNN on CESNET-QUIC22 with the IETF QUIC v1 protocol, and a 1.09% improvement on CipherSpectrum with TLS 1.3 for 40-class website fingerprinting.

Förderung

  • Defense Research and Development Organization (DRDO), Ministry of Defense, Government of India

Zitieren

APA

Dutta, A., Anand, P., Roy, S., & Chakraborty, R. S. (2026). KAN-Vis: Efficient and Lightweight Visual Technique for Network Traffic Classification Using Kolmogorov–Arnold Network. IEEE Transactions on Networking, 34, 6595-6605. https://doi.org/10.1109/ton.2026.3722338

Vancouver

Dutta A, Anand P, Roy S, Chakraborty RS. KAN-Vis: Efficient and Lightweight Visual Technique for Network Traffic Classification Using Kolmogorov–Arnold Network. IEEE Transactions on Networking. 2026;34:6595-6605. doi:10.1109/ton.2026.3722338

BibTeX
@article{Dutta2026,
  author = {Dutta, Anurag and Anand, Pallavi and Roy, Sangita and Chakraborty, Rajat Subhra},
  title = {KAN-Vis: Efficient and Lightweight Visual Technique for Network Traffic Classification Using Kolmogorov–Arnold Network},
  journal = {IEEE Transactions on Networking},
  year = {2026},
  volume = {34},
  pages = {6595--6605},
  doi = {10.1109/ton.2026.3722338}
}
Kostenlos starten →