System and method of classifying network traffic
Patent Information
- Application Number
- EP2024763365
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-02-27
- Filing Date
- 2024-02-26
- Publication Date
- 2026-01-07
AI Technical Summary
Current network traffic classification methods, particularly those using Deep Packet Inspection (DPI) tools, are ineffective with encrypted traffic, leading to decreased reliability and inefficiency, especially with the upcoming encryption standards like DNS over HTTPS, which complicates the classification of metadata essential for network traffic analysis.
A system and method utilizing a combination of parametric and nonparametric machine-learning (ML) models to classify network traffic by capturing and sampling network data, calculating vector embeddings in a feature space, and applying ML-based models to classify traffic based on similarity and confidence metrics, even with encrypted data, allowing for reliable classification of encrypted network traffic.
The approach enhances the reliability and efficiency of network traffic classification by reducing computational loads and adapting to evolving network traffic patterns, providing accurate classification of encrypted traffic and improving the scalability and adaptability of classification systems.
Smart Images
Figure IL2024050220_06092024_PF_FP