Self-Supervised One-Class Tabular Deep Learning for Intrusion Detection in Artificial Intelligence of Things Systems

Abstract

The emergence of the artificial intelligence of things (AIoT) has led to increasingly sophisticated cyberattacks. For many of these threats, labeled training data are scarce or unavailable, which increases the need for label-free intrusion detection systems. To address this issue, this study proposes a self-supervised one-class TabNet framework, termed SS-OC-TabNet, for intrusion detection in tabular network traffic. The framework combines self-supervised representation learning and one-class compactness learning within a two-stage training process. In the first stage, masked-feature reconstruction is used to learn the feature structure of normal traffic from unlabeled data. In the second stage, a hypersphere compactness objective is applied to make the latent representations of normal traffic more compact for anomaly detection. TabNet is used as the backbone to support sequential feature selection, which also improves interpretability. The proposed model was evaluated on the NSL-KDD dataset and further validated through comparison with related works on the ToN-IoT dataset, achieving an accuracy of 93.20% with a false alarm rate of 2.13%. Further, the interpretability results show which traffic features influence the model's decisions. The compact design and low computational complexity of the proposed framework also support energy-efficient deployment in resource-constrained AIoT environments. These results indicate that SS-OC-TabNet is a viable tool for label-free intrusion detection in AIoT environments.


Received: 17 October 2025 | Revised: 2 July 2026 | Accepted: 10 September 2026



Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.



Data Availability Statement

The KDD CUP 99 datasets that support the findings of this study are openly available at https://doi.org/10.1109/CISDA.2009.5356528, reference number [35] and the Network TON_IoT datasets are openly available at https://doi.org/10.1016/j.scs.2021.102994, reference number [36].



Author Contribution Statement

Thavavel Vaiyapuri: Conceptualization, Methodology, Formal analysis, Writing – original draft, Supervision, Funding acquisition. Karthiyayini Murugesan: Methodology, Software, Resources, Writing – original draft. Yassine Bouteraa: Software, Validation, Formal analysis, Investigation, Resources, Writing – review & editing, Visualization, Project administration.