Network Traffic Categorization via Embedding Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for determining network traffic categories face challenges such as scalability issues, adaptability problems due to evolving network patterns, privacy concerns, and difficulties with encrypted traffic, leading to inefficient resource utilization and incorrect network modifications.
Innovation Solution
A method utilizing machine learning models to categorize network traffic by transforming packet size data into embeddings, generating similarity metrics, creating graphs, and applying community detection models to identify traffic categories without prior knowledge of application categories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual catalogs or prior domain knowledge are used to determine network traffic categories, then network traffic categorization can be performed, but scalability issues arise and the system cannot adapt to evolving network patterns
Solution Approach 1:
The system performs self-service by automatically generating network traffic categories through machine learning models without requiring manual catalog creation or maintenance. The models process network traffic data, generate embeddings, and automatically identify categories, enabling the system to adapt to evolving patterns autonomously
Solution Approach 2:
The system changes parameters by transforming network traffic data into embeddings through machine learning models, converting raw traffic characteristics into a format that reveals categorical patterns. This parameter transformation enables automatic category identification without manual intervention
2Reliability
If traditional network traffic categorization methods are used, then some level of categorization is achieved, but resource utilization becomes inefficient and incorrect network modifications occur
Solution Approach 1:
The system replaces traditional mechanical categorization methods with machine learning-based automatic classification. Instead of manual rule-based systems, the patent uses neural networks and embedding models to substitute and achieve more accurate categorization with optimized resource usage
3Measurement precision
If application information is used to determine network traffic categories, then categorization accuracy improves, but privacy concerns arise and encrypted traffic cannot be handled
Solution Approach 1:
The system extracts categorical information from network traffic data without extracting or requiring application-level information. By taking out only the necessary traffic pattern features and processing them through machine learning models, the system achieves accurate categorization while preserving privacy and working with encrypted traffic
Data Source
AI summary
A device may receive network traffic data that includes network traffic packet sizes, and may transform the network traffic data into transformed data. The device may process the transformed data, with a machine learning model, to generate an embedding, and may obtain a similarity metric for the embedding. The device may create a graph with nodes and edges based on the embedding and the similarity metric, and may process the graph, with a community detection model, to identify network traffic categories for the network traffic data. The device may perform one or more actions based on the network traffic categories.