MODEL-AGNOSTIC ANALYSIS OF BGP FEATURE DYNAMICS FOR ANOMALY DETECTION IN INTERDOMAIN ROUTING
Information and communication technologies, electronic engineering · 2026-06 · 6 (1) · 1-10
DOI: 10.23939/ictee2026.01.001 ↗Kurz gefasst
Fachartikel in der Zeitschrift Information and communication technologies, electronic engineering, 2026-04. Verfasst von S. Maruniak und 1 weiteren Autor:innen. Die Arbeit nennt 16 Literaturangaben.
Abstract (Originalsprache)
Anomaly detection in interdomain routing commonly relies on numerical features derived from Border Gateway Protocol (BGP) updates, yet the temporal behavior of these features is usually assessed only through the performance of predictive models. This limits understanding of how routing characteristics evolve throughout an incident and whether particular descriptors consistently reflect abnormal conditions. This paper presents a model-agnostic method for analyzing feature dynamics directly from observational data. Each event is represented by a time interval divided into pre-anomaly, anomaly, and post-anomaly phases, enabling explicit comparison between baseline activity, disturbance, and recovery. For every phase, established statistical measures describing variability, distributional change and separability are computed. Based on these values, normalized coefficients are introduced to quantify the magnitude of anomaly response and the degree of return toward baseline, allowing comparison across diverse features without dependence on scale or metric type. A practical criterion is further defined to identify phase-informative descriptors that exhibit both pronounced deviation during the anomaly and subsequent stabilization. The method was evaluated on a set of documented routing incidents of different types using time-series features extracted from public measurements. The results indicate that only a limited subset of descriptors consistently demonstrates strong contrast between normal and abnormal phases, while many averaged structural and temporal indicators remain weakly responsive. The proposed approach enables systematic assessment of feature sets prior to model construction, supports comparison of alternative extraction strategies, assists in identifying descriptors that contribute meaningful information about abnormal routing conditions, and provides model-independent explanatory insight that can complement explainable artificial intelligence techniques (XAI) and improve interpretability of subsequent detection models.
Zitieren
APA
Maruniak, S., & Kyryk, M. (2026). MODEL-AGNOSTIC ANALYSIS OF BGP FEATURE DYNAMICS FOR ANOMALY DETECTION IN INTERDOMAIN ROUTING. Information and communication technologies, electronic engineering, 6(1), 1-10. https://doi.org/10.23939/ictee2026.01.001
Vancouver
Maruniak S, Kyryk M. MODEL-AGNOSTIC ANALYSIS OF BGP FEATURE DYNAMICS FOR ANOMALY DETECTION IN INTERDOMAIN ROUTING. Information and communication technologies, electronic engineering. 2026;6(1):1-10. doi:10.23939/ictee2026.01.001
BibTeX
@article{Maruniak2026,
author = {Maruniak, S. and Kyryk, M.},
title = {MODEL-AGNOSTIC ANALYSIS OF BGP FEATURE DYNAMICS FOR ANOMALY DETECTION IN INTERDOMAIN ROUTING},
journal = {Information and communication technologies, electronic engineering},
year = {2026},
volume = {6},
number = {1},
pages = {1--10},
doi = {10.23939/ictee2026.01.001}
}
EUVIMED