Data Packet Flow Classification with Confidence Degree
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data packet flow classification techniques in communication networks are prone to errors and uncertainties, leading to unreliable operation and inappropriate policy applications, which can degrade network reliability.
Innovation Solution
The proposed solution involves classifying data packet flows with a confidence degree by using a hierarchical structure of identifiers (flow, protocol, service, subscriber, and network) and incorporating a confidence level value to indicate the trustworthiness of these identifiers, which can be determined through various inspection techniques such as Deep Packet Inspection (DPI) and Shallow Packet Inspection (SPI).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data packet flow classification is performed using prior-art techniques, then classification can be achieved, but classification reliability is degraded due to errors and uncertainties
Solution Approach 1:
The patent implements a feedback mechanism where the confidence level value is continuously updated based on the inspection results of data packets. The classification reliability is improved by using the confidence level as feedback to determine whether to apply a policy or to perform further inspection, creating a closed-loop system that adapts to the actual classification certainty.
Solution Approach 2:
The patent performs preliminary inspection of data packets to determine a confidence level value before applying classification policies. By conducting preliminary actions (inspection) to assess classification certainty, the system avoids making policy decisions based on uncertain classifications, thereby improving overall reliability.
2Device complexity
If classification is performed without confidence degree information, then processing is simpler, but policy application becomes inappropriate and network operation reliability degrades
Solution Approach 1:
The patent segments the classification system into distinct components: packet inspection, confidence level determination, classification decision, and policy application. By adding the confidence level determination as a separate segment, the system maintains modularity while improving reliability through informed decision-making.
3Measurement precision
If Deep Packet Inspection (DPI) techniques are used to improve classification accuracy, then measurement precision improves, but processing time and computational complexity increase
Solution Approach 1:
The patent implements a dynamic inspection approach where the inspection depth and method are adjusted based on the determined confidence level. When shallow inspection provides sufficient confidence, further deep inspection is avoided, thus reducing time loss while maintaining precision when necessary.
Solution Approach 2:
The patent changes the inspection parameter (confidence level value) based on the results of preliminary inspection. By dynamically adjusting the inspection thoroughness according to the confidence level, the system optimizes the trade-off between classification precision and processing time.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
At least one identifier classifies a data packet flow (110) of data packets (115) between an originator (301) and the recipient (302). A confidence level value of the at least one identifier specifies a confidence degree with which the at least one identifier is determined.