![[2602.19785] Unsupervised Anomaly Detection in NSL-KDD Using $β$-VAE: A Latent Space and Reconstruction Error Approach](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19785] Unsupervised Anomaly Detection in NSL-KDD Using $β$-VAE: A Latent Space and Reconstruction Error Approach
Summary
This paper presents an unsupervised anomaly detection method using β-VAE on the NSL-KDD dataset, comparing latent space structure and reconstruction error approaches.
Why It Matters
As cybersecurity threats evolve, effective anomaly detection in network traffic is crucial. This research provides insights into advanced methodologies that enhance the performance of Intrusion Detection Systems, making it relevant for both academia and industry.
Key Takeaways
- Introduces a novel unsupervised anomaly detection method using β-VAE.
- Compares the effectiveness of latent space structure and reconstruction error metrics.
- Demonstrates the advantages of latent space exploitation for classification tasks.
- Highlights the importance of effective anomaly detection in cybersecurity.
- Provides experimental results that validate the proposed approaches.
Computer Science > Machine Learning arXiv:2602.19785 (cs) [Submitted on 23 Feb 2026] Title:Unsupervised Anomaly Detection in NSL-KDD Using $β$-VAE: A Latent Space and Reconstruction Error Approach Authors:Dylan Baptiste (CRESTIC), Ramla Saddem (CRESTIC), Alexandre Philippot (CRESTIC), François Foyer View a PDF of the paper titled Unsupervised Anomaly Detection in NSL-KDD Using $\beta$-VAE: A Latent Space and Reconstruction Error Approach, by Dylan Baptiste (CRESTIC) and 3 other authors View PDF Abstract:As Operational Technology increasingly integrates with Information Technology, the need for Intrusion Detection Systems becomes more important. This paper explores an unsupervised approach to anomaly detection in network traffic using $\beta$-Variational Autoencoders on the NSL-KDD dataset. We investigate two methods: leveraging the latent space structure by measuring distances from test samples to the training data projections, and using the reconstruction error as a conventional anomaly detection metric. By comparing these approaches, we provide insights into their respective advantages and limitations in an unsupervised setting. Experimental results highlight the effectiveness of latent space exploitation for classification tasks. Subjects: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Machine Learning (stat.ML) Cite as: arXiv:2602.19785 [cs.LG] (or arXiv:2602.19785v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2602.19785 Focus to ...