Related Experiment Videos
FET-FIDS: a federated enhanced transformer-based framework for privacy-preserving network intrusion detection
Jothi Prabha Appadurai1, Revoori Swetha2, Vasam Srinivas3
1Department of CSE (AI&ML), Kakatiya Institute of Technology and Science, Warangal, India.
Scientific Reports
|July 8, 2026
Summary
A new Federated Enhanced Transformer-based Intrusion Detection System (FET-IDS) enhances cybersecurity for interconnected systems. This privacy-preserving approach improves intrusion detection accuracy and scalability in distributed environments.
Area of Science:
- Cybersecurity and Network Engineering
- Artificial Intelligence and Machine Learning
Background:
- Modern digital infrastructures face increasing cyber threats due to interconnected systems and distributed architectures.
- Traditional Intrusion Detection Systems (IDS) struggle with privacy, scalability, and single points of failure in large networks.
- Existing federated learning IDS models face challenges with convergence, non-IID data, and high computational costs.
Purpose of the Study:
- To propose a novel Federated Enhanced Transformer-based Intrusion Detection System (FET-IDS) for privacy-preserving and decentralized network security.
- To address the limitations of existing IDS, including scalability, convergence speed, and handling of non-IID data.
- To enhance intrusion detection capabilities in distributed and heterogeneous network environments.
Main Methods:
- Implemented a decentralized security system combining federated learning with Transformer-based self-attention.
- Trained local FET-IDS models on client network traffic data for pattern learning.
- Utilized adaptive federated averaging on a centralized server to combine model updates securely, preserving data confidentiality.
Main Results:
- The proposed FET-IDS demonstrated improved scalability, robustness, and communication performance in distributed environments.
- Achieved a high detection accuracy of 97.82% under experimental conditions.
- Showcased superior effectiveness in detection, stability, and convergence compared to existing IDS models.
- Effectively handled non-IID data distributions and proved extensible to various network environments.
Conclusions:
- FET-IDS offers a robust, privacy-preserving, and decentralized solution for intrusion detection in complex networks.
- The Transformer-based approach enhances the ability to learn intricate intrusion patterns.
- The system effectively overcomes the limitations of traditional and existing federated IDS, particularly in heterogeneous and distributed settings.
