adversarial-representation-learning

This paper proposes a zero-shot learning framework for multivariate IoT traffic anomaly detection using adversarial and contrastive learning within a variational autoencoder, enabling domain adaptation without labeled data and demonstrating strong performance across diverse datasets.