Van-Tam Nguyen

We propose a self-supervised framework that combines GNNs and a Transformer-based masked autoencoder to detect network intrusions by reconstructing flow representations and flagging high-error patterns as anomalies.