TabNet-IDS: A TabNet-Driven Tabular Deep Learning Framework for Intrusion Detection Systems

The rapid growth of networked systems has increased exposure to sophisticated cyberattacks, demanding intrusion detection methods that can learn directly from complex traffic data. This paper proposes a TabNet-driven intrusion detection framework built on the UNSW-NB15 benchmark dataset. The approach first performs data cleaning, label encoding of categorical attributes, and standardization of numerical features, followed by a binary labeling scheme that groups all attack types into a single intrusion class against normal traffic. The TabNet classifier is trained using 75% of the labeled data, while 25% is reserved for validation, and its generalization is assessed on an independent unseen test set. Experimental results show strong and balanced performance, achieving a precision of 92.68%, recall of 94.23%, F1-score of 93.45%, accuracy of 92.72%, and a Matthews Correlation Coefficient of 85.28%. The ROC and Precision–Recall curves further confirm high separability between normal and intrusive flows, with areas under the curve of 0.975 and 0.977, respectively. A comparison with several representative state-of-the-art models, including GRU-based, decision tree, ensemble, and geometric-analysis methods, demonstrates that the proposed TabNet framework provides the most favorable combination of evaluation metrics, indicating its suitability for accurate and reliable network intrusion detection.

• Publication year

  2026

• Venue

  International Conference on Applied Informatics and Communication

• Publication date

  2026-02-18

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/ICAIC67076.2026.11395861
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Intelligent network intrusion detection system using optimized deep CNN-LSTM with UNSW-NB15

  2025cited by this paper
• An intrusion detection system for industrial IoT using CNN-LSTM and reptile search algorithm

  2025cited by this paper
• Development of an intelligent intrusion detection system for IoT networks using deep learning

  2025cited by this paper
• Smart deep learning model for enhanced IoT intrusion detection

  2025influential reference
• A network intrusion detection system based on deep learning in the IoT

  2024cited by this paper
• Enhanced Intrusion Detection Systems Performance with UNSW-NB15 Data Analysis

  2024cited by this paper
• Advancing Intrusion Detection with Machine Learning: Insights from the UNSW-NB15 Dataset

  2024cited by this paper
• Optimizing IoT Network Intrusion Detection: A Deep Learning Approach

  2024cited by this paper
• Improved convolution neural network integrating attention based deep sparse auto encoder for network intrusion detection

  2024cited by this paper
• A new two-phase intrusion detection system with Naïve Bayes machine learning for data classification and elliptic envelop method for anomaly detection

  2023cited by this paper
• A deep learning technique for intrusion detection system using a Recurrent Neural Networks based framework

  2022influential reference
• ToN_IoT: The Role of Heterogeneity and the Need for Standardization of Features and Attack Types in IoT Network Intrusion Data Sets

  2021cited by this paper
• Novel Geometric Area Analysis Technique for Anomaly Detection Using Trapezoidal Area Estimation on Large-Scale Networks

  2019influential reference
• The evaluation of Network Anomaly Detection Systems: Statistical analysis of the UNSW-NB15 data set and the comparison with the KDD99 data set

  2016cited by this paper
• UNSW-NB15: a comprehensive data set for network intrusion detection systems (UNSW-NB15 network data set)

  2015cited by this paper

• No citing papers are available for this paper.