عنوان مقاله:

نویسندگان:

Mahsa Raeiszadeh، Hooman Tahayori

The complexity and ever-changing nature of mi- croservice architectures pose significant challenges for ensuring system reliability and availability. As such, efficient anomaly detection is vital for averting potential failures and resolving issues promptly within microservice systems. Recently, distributed data analysis methods powered by Artificial Intelligence (AI) have been increasingly recognized for their effectiveness in detecting anomalies in these environments. However, AI-based anomaly detection techniques often depend on centralized data aggregation and processing, which can lead to scalability concerns. In this paper, we introduce a Federated Learning (FL)-based distributed collaborative AI approach designed to detect anomalies in mi- croservice systems. The proposed solution captures intra-service behaviors and inter-service dependencies using a Span Causal Graph (SCG) to represent multi-source distributed data and com- bines a Graph Neural Network (GNN) with PU learning for model training. This approach is integrated into the FL framework to facilitate scalable and accurate anomaly detection. Trace-driven evaluations show that the proposed approach surpasses existing state-of-the-art methods in terms of accuracy, while also satisfying time efficiency and scalability requirements

Anomaly Detection, Cloud, Federated Learning, Microservices, Tracing data.

برای لینک‌دهی به این مقاله از آدرس زیر استفاده نمایید:

برای ارجاع در منابع از عبارت زیر استفاده کنید:
Raeiszadeh، Mahsa، Tahayori، Hooman، 1403، Federated Learning with Graph Trace Analysis for Anomaly Detection in Cloud Microservices، دومین کنفرانس بین المللی هوش مصنوعی و مهندسی نرم افزار، https://conf.shirazu.ac.ir/fa/archive_index.php?c=HN-aisoft24&aid=1509278