SMARTY: Confidentiality-Preserving Real-Time Localization of Soft Failures in Optical Networks
This research, available through the SMARTY community on Zenodo, presents an approach to localizing soft failures in optical networks while preserving data confidentiality through dimensionality reduction and machine learning.
Soft failures are gradual degradations in network performance that can affect service reliability before a complete failure occurs. Their early detection and localization are essential for ensuring reliable optical communications. However, using external machine learning services to analyze network data raises concerns about information confidentiality.

The proposed solution combines Principal Component Analysis (PCA) with Machine Learning as a Service (MLaaS). PCA reduces the dimensionality of network telemetry data before sharing it with an external service, which uses semi-supervised K-means clustering to identify network conditions and locate potential failures.
The method processes new network observations every four seconds and achieves localization accuracy of up to 98.5% on large-scale simulated datasets and 98% on experimental data, demonstrating its effectiveness in combining accurate monitoring with data confidentiality.
Read the publication: https://zenodo.org/records/22095663
Yeganehfallah, A., Sgambelluri, A., Paolini, E., Mayer, K. S., Silva, M. F., Mello, D. A. A., & Valcarenghi, L., “Confidentiality-preserving real-time localization of soft failures in optical networks based on PCA and MLaaS,” Journal of Optical Communications and Networking, vol. 17, no. 9, pp. D137–D143, 2025, doi: 10.1364/JOCN.562802.
Follow SMARTY for updates:
LinkedIn: https://www.linkedin.com/company/smartyeu/?viewAsMember=true
Website: https://www.smarty-project.eu
