DUdetector is a dual-granularity unsupervised model for network anomaly detection, utilizing an Ehanced Transformer and a Conv1d&MaxPool1d AutoEncoder with residual connections (abbr., CM&RC-AE) to perform both coarse-grained segment-level and fine-grained point-level detection.
This repo is part of our work:
DUdetector: A Dual-granularity Unsupervised Model for Network Anomaly Detection
If you use the source code, please cite our paper.