EGNN: energy-efficient anomaly detection for IoT multivariate time series data using graph neural network - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Future Generation Computer Systems Année : 2024

EGNN: energy-efficient anomaly detection for IoT multivariate time series data using graph neural network

Résumé

Due to the rapid growth of multiple application areas of the Internet of Things (IoT), including transportation, secure water treatment, smart grids, and smart homes, the number of IoT devices is increasing. These IoT devices generate a large amount of sensory data during operation at the network edge [1], most of which are multivariate time series [2]. Detecting anomalies upon these data has been applied to guarantee the health of IoT applications, such as intelligent transportation and autonomous driving [3], [4]. Traditional techniques have been proposed to detect various types of anomalies with different measurement design and discovery algorithms on the cloud [5], [6]. At the same time, the increasing use of battery-powered IoT devices in edge scenarios poses a significant challenge due to limited rechargeability [7]. This emphasizes the critical need for accurate and energy-efficient anomaly detection algorithms in IoT environments. However, such methods still incur a large amount of energy consumption for data transmission. However, integrating such algorithms into IoT systems can lead to a substantial increase in energy consumption for data transmission, resulting in higher operational costs. To tackle these limitations, this paper aims to propose a novel energy-saving anomaly detection algorithm specifically designed for IoT deployments.
Fichier non déposé

Dates et versions

hal-04376917 , version 1 (07-01-2024)

Identifiants

Citer

Hongtai Guo, Zhangbing Zhou, Deng Zhao, Walid Gaaloul. EGNN: energy-efficient anomaly detection for IoT multivariate time series data using graph neural network. Future Generation Computer Systems, 2024, 151, pp.45-56. ⟨10.1016/J.FUTURE.2023.09.028⟩. ⟨hal-04376917⟩
31 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More