Experimental Analysis of Adaptive ML Classifiers for Dynamic Detection of Emerging Physical-Layer Attacks
Paper i proceeding, 2025
We experimentally evaluate three ML classifiers for detecting physical-layer attacks in optical networks. Using telemetry data from a real-world testbed under six attack types, we achieve Balanced Accuracy of up to 0.936 for unseen attacks, and demonstrate rapid adaptation of Multilayer Perceptron to evolving threats.
data stream
machine learning
attack detection
Författare
Aleksandra Knapinska
Politechnika Wrocławska
Chalmers, Elektroteknik, Kommunikation, Antenner och Optiska Nätverk
Marija Furdek Prekratic
Chalmers, Elektroteknik, Kommunikation, Antenner och Optiska Nätverk
European Conference on Optical Communication, ECOC
9798331595319 (ISBN)
51st European Conference on Optical Communication (ECOC 2025)
Copenhagen, Denmark,
Copenhagen, Denmark,
5G betrodda och säkra nätverkstjänster (5G-TACTIC)
Europeiska kommissionen (EU) (EC/H2020/101127973), 2023-12-01 -- 2026-11-30.
Styrkeområden
Informations- och kommunikationsteknik
Ämneskategorier (SSIF 2025)
Kommunikationssystem
Datavetenskap (datalogi)
Datorsystem
DOI
10.1109/ECOC66593.2025.11263047
ISBN
9798331595319