An asynchronous federated learning network intrusion detection method based on similarity gating
By introducing an asynchronous federated learning method with feature pre-screening, similarity gating, and staleness decay mechanism on both the client and server sides, the problems of data privacy leakage, staleness updates, and computational burden in the Industrial Internet of Things are solved, and efficient network intrusion detection is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XIJING UNIV
- Filing Date
- 2026-04-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing network intrusion detection methods suffer from several problems in distributed scenarios such as the Industrial Internet of Things (IIoT), including the risk of privacy leaks due to centralized uploading of raw data, difficulty in effectively filtering out outdated and low-quality updates during asynchronous aggregation, and heavy computational burden on the edge due to high-dimensional network traffic characteristics.
An asynchronous federated learning method based on similarity gating is adopted. By introducing feature pre-screening and feature search mechanism under federated constraints on the client side, high-dimensional network traffic features are optimized. On the central server side, similarity gating and staleness decay mechanism are introduced to filter asynchronously uploaded updates. An asynchronous weighted aggregation strategy combining buffer batch threshold and timeout threshold is adopted to achieve flexible aggregation.
It reduces the computational and storage burden on the edge, improves the stability and detection reliability of asynchronous federated aggregation, enhances the application adaptability in multi-domain distributed scenarios, and reduces communication overhead and training latency.
Smart Images
Figure CN122419862A_ABST