A Unified Siamese Learning Framework for Zero-Day Anomaly Detection and Classification in Optical Networks

2026-06-09

AI Summary

This work addresses the challenge of detecting and classifying zero-day anomalies—previously unseen anomaly types—in emerging optical network paths. To this end, the authors propose a unified Siamese neural network framework based on multiple similarity metrics that jointly tackles zero-day anomaly detection and one-shot classification. By modeling the relative relationships between normal and anomalous patterns in the feature space, the method generalizes to unknown anomalies and new optical paths without requiring model retraining. Experimental results demonstrate that the proposed approach achieves over 99% accuracy across diverse optical network scenarios, marking the first solution capable of immediate adaptation without retraining. This capability significantly enhances the timeliness and robustness of anomaly response in optical networks.

Abstract

A multi-similarity Siamese neural network unifies zero-day anomaly detection and one-shot classification in optical networks, achieving over 99% accuracy and instant adaptability across lightpaths and unseen anomaly types without any retraining.

Problem

• zero-day anomaly
• anomaly detection
• anomaly classification
• optical networks
• one-shot learning

Innovation

• Siamese neural network
• zero-day anomaly detection
• one-shot classification
• optical networks
• multi-similarity learning