Enhanced training of response time anomaly detectors using diffusion models
File(s)
Accepted version
Author(s)
Luo, Wenxiang
Casale, Giuliano
Type
Conference Paper
Abstract
Machine learning (ML) approaches have grown in popularity in recent years as a way to identify anomalies in microservice-based applications. However, training ML models may suffer from time constraints and limited real-world failure data availability. In this paper, we propose a method to mitigate this issue using diffusion models for training data augmentation. Response time data collected using distributed traces is used to train diffusion models leveraging a customized UNet proposed in the paper. The resulting diffusion models can then generate new data to improve the training of response time anomaly detectors. Experiments using the DeathStarBench microservices architecture demonstrate that the proposed approach increases the accuracy after training of anomaly detection models by 20%. We further show that the response time data generated by our diffusion models cannot be distinguished by classic discriminators, which confirms that the generated data are of high quality.
Date Issued
2025-12-21
Date Acceptance
2025-08-05
Citation
2025 33rd International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS), 2025, pp.1-8
Publisher
IEEE
Start Page
1
End Page
8
Journal / Book Title
2025 33rd International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS)
Copyright Statement
© 2025 IEEE. This is the author’s accepted manuscript made available under a CC-BY licence in accordance with Imperial’s Research Publications Open Access policy (www.imperial.ac.uk/oa-policy)
License URL
Source
MASCOTS 2025
Publication Status
Published
Start Date
2025-10-21
Finish Date
2025-10-23
Coverage Spatial
Paris, France