LoGREAT: An Efficient Log Anomaly Detection Model for 5G RedCap Power Distribution Networks in Fading FBL Regime
IEEE Communications Letters · 2026 · 30 · 2590-2594
DOI: 10.1109/lcomm.2026.3712608 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Communications Letters, 2026. Verfasst von Yuming Han und 1 weiteren Autor:innen. Die Arbeit nennt 17 Literaturangaben.
Abstract (Originalsprache)
Current intelligent power distribution networks face the challenge of detecting anomaly logs and transmitting them efficiently and reliably over fading 5G short-packet channels. To address this, we first develop a finite-blocklength analysis for short-packet transmission in fading 5G networks, revealing the gap between capacity-based design and practical achievable performance. Based on this, we introduce an importance weight to prioritize logs by their anomaly significance. We then propose LoGREAT, an unsupervised Transformer–CNN framework that captures both global temporal dependencies and local burst patterns. To further improve transmission efficiency, we design a Deep Q-Learning (DQN) based decision making module that combines anomaly scores with channel state information to transmit logs. Experimental results demonstrate the effectiveness of the proposed framework for reliable and intelligent anomaly-aware communication.
Zitieren
APA
Han, Y., & Kang, J. (2026). LoGREAT: An Efficient Log Anomaly Detection Model for 5G RedCap Power Distribution Networks in Fading FBL Regime. IEEE Communications Letters, 30, 2590-2594. https://doi.org/10.1109/lcomm.2026.3712608
Vancouver
Han Y, Kang J. LoGREAT: An Efficient Log Anomaly Detection Model for 5G RedCap Power Distribution Networks in Fading FBL Regime. IEEE Communications Letters. 2026;30:2590-2594. doi:10.1109/lcomm.2026.3712608
BibTeX
@article{Han2026,
author = {Han, Yuming and Kang, Jiawen},
title = {LoGREAT: An Efficient Log Anomaly Detection Model for 5G RedCap Power Distribution Networks in Fading FBL Regime},
journal = {IEEE Communications Letters},
year = {2026},
volume = {30},
pages = {2590--2594},
doi = {10.1109/lcomm.2026.3712608}
}
EUVIMED