Ensemble Learning for Network Intrusion Detection using FT-Transformer and Traditional Learning Models
- Version
- Download 8
- File Size 461.67 KB
- File Count 1
- Create Date 27 January 2026
- Last Updated 27 January 2026
Ensemble Learning for Network Intrusion Detection using FT-Transformer and Traditional Learning Models
Jeewan Kumar Thakur
Cloud Technology and Information Security
Jain University Banglore, India
21btrci012@jainuniversity.ac.in
Dipendra kumar Singh
Cloud Technology and Information Security
Jain University Banglore, India
21btrci009@jainuniversity.ac.in
Devashish Prajapat
Cloud Technology and Information Security
Jain University Banglore, India
21btrci008@jainuniversity.ac.in
Dr. Sonal Sharma
Cloud Technology and Information Security(HOD) Jain University(Banglore, India) s.sonal@jainuniversity.ac.in
Abstract— Information system defense requires network intrusion detection, especially in situations that mimic military network activities. This study offers a comprehensive strategy to increase the accuracy of network intrusion detection by combining traditional machine learning techniques with advanced transformer-based designs. We perform an analysis using a number of models, such as Support Vector Machines (SVM), Logistic Regression (LR), K-Nearest Neighbors (KNN), Decision Trees, Random Forests, and a novel Feature Tokenizer Transformer (FT-Transformer), on a dataset that was extracted from a simulated US Air Force LAN environment. Each model was thoroughly trained and tested to predict and distinguish between normal and aberrant TCP/IP connections. The FT- Transformer demonstrated a significant improvement in detection performance by using feature tokenization tailored for tabular data, achieving 99.78% accuracy and 99.75% recall in identifying attack paths from typical data. Comparisons of evaluations show that the hybrid ensemble approach produced a consistent outcome by combining the output of many models, increasing the estimated accuracy to 99.78%. The findings show the benefits of combining ensemble artificial intelligence techniques with transformer designs for network intrusion detection, paving the way for more intelligent and robust cybersecurity systems.
Keywords— Network Intrusion Detection, FT- Transformer, SVM, LR, KNN, Decision Tree Random Forest, Voting Classifier.
Download