Data Stream Identification via Feature Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The accuracy of identifying application identifiers in data streams is low due to encrypted payload fields and private application layer protocols, which complicates feature extraction and matching with existing methods.
Innovation Solution
A data stream identification method that generates feature images from packet transmission attribute information of consecutive packets, using static and dynamic behavior feature matrices, and inputs these into a pre-trained image classification model to determine the application identifier, avoiding payload field parsing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If deep packet inspection method is used to match payload fields with feature library, then identification process is straightforward, but accuracy deteriorates when payload fields are encrypted or use private protocols
Solution Approach 1:
The patent introduces packet transmission attribute information as an intermediary between the data stream and the identification system. Instead of directly analyzing the payload field (which fails when encrypted), the system uses transmission attributes (packet length, arrival time, inter-arrival time) as a mediator to infer application identity, thereby resolving the contradiction between ease of identification and accuracy under encryption.
Solution Approach 2:
The patent replaces the mechanical payload field parsing approach with a statistical analysis approach. Instead of manually extracting and matching payload content (which breaks down with encryption), the system substitutes this with collecting and analyzing transmission attribute statistics, achieving accurate identification without payload decryption.
2Measurement precision
If payload field parsing is performed to extract feature fields, then feature extraction is direct, but user privacy is invaded and accuracy drops when encryption is used
Solution Approach 1:
The patent extracts only the necessary transmission attribute information (packet length, arrival time, direction) from the packet headers, completely avoiding extraction of payload field content. This selective extraction approach maintains identification accuracy while preserving user privacy by never touching the encrypted payload data.
Solution Approach 2:
The transmission attributes serve as an intermediary that allows the system to identify applications without accessing the payload field. This mediator approach enables accurate feature extraction for identification purposes while completely avoiding privacy invasion, as the intermediary operates on public header information only.
3Measurement precision
If multiple packet transmission attributes are collected and processed into feature images, then identification accuracy improves, but processing complexity increases
Solution Approach 1:
The patent segments the packet transmission attributes into distinct categories (packet length, arrival time, inter-arrival time, direction) and processes each separately to generate corresponding feature images. This segmentation allows complex multi-dimensional data to be handled through simpler, modular processing steps, reducing overall system complexity while maintaining high accuracy.
Solution Approach 2:
The patent transforms various packet transmission attributes into a standardized feature image format through parameter transformation. By converting different attribute types (time intervals, lengths, directions) into a unified visual feature representation, the system simplifies the processing complexity and enables use with standard image classification models.
Data Source
AI summary
This application provides a data stream identification method and apparatus and belongs to the field of Internet technologies. The method includes: obtaining packet transmission attribute information of N consecutive packets in a target data stream; generating feature images of the packet transmission attribute information of the N consecutive packets based on the packet transmission attribute information of the N consecutive packets; and inputting the feature images into a pre-trained image classification model, to obtain a target application identifier corresponding to the target data stream. According to this application, accuracy of identifying an application identifier corresponding to a data stream can be improved.