Adaptive Statistical Traffic Classification in Broadband Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data traffic classification methods in broadband networks are inaccurate and inefficient, failing to effectively prioritize and manage diverse data types such as VOIP, video streaming, and web browsing due to limitations in IP and port number classification, deep packet inspection, and statistical methods that are subjective and prone to changes in traffic patterns.
Innovation Solution
A rule-based statistical classification method that adaptively tunes classification based on network conditions, using statistical factors like throughput rate, session duration, and inter-arrival times to accurately classify data traffic sessions into priority types, ensuring efficient resource allocation and quality of service (QoS) across different data types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If IP address and port number classification is used, then classification speed is improved, but classification accuracy deteriorates because the same port may be used for different data types with different transmission requirements
Solution Approach 1:
The patent combines multiple classification approaches (IP/port number classification, statistical inference, and deep packet inspection) into a unified hybrid classification system. This merging allows the system to leverage the speed of IP/port classification while compensating for its accuracy limitations through statistical analysis and selective DPI, thereby achieving both high classification speed and high accuracy simultaneously.
2Measurement precision
If deep packet inspection is used, then classification accuracy is improved, but processing overhead and complexity increase significantly especially as more Internet data traffic is secured
Solution Approach 1:
The patent segments the traffic classification process into multiple stages: first using IP/port number classification for quick initial categorization, then applying statistical inference to further differentiate traffic types, and finally using deep packet inspection only for specific cases where earlier methods are inconclusive. This segmentation reduces the overall processing complexity by avoiding comprehensive DPI for all traffic while maintaining high classification accuracy.
Solution Approach 2:
The patent applies deep packet inspection partially rather than excessively - only for traffic that requires higher accuracy and cannot be classified by lighter methods. This partial application of DPI reduces processing overhead and complexity while still achieving the necessary classification accuracy for secured and complex traffic types.
3Adaptability or versatility
If statistical classification methods are used, then adaptability to traffic patterns is improved, but classification accuracy deteriorates because the methods are blind and subject to changes in traffic patterns
Solution Approach 1:
The patent incorporates feedback mechanisms where classification results and network conditions are continuously monitored and used to adjust classification parameters and thresholds. This feedback loop enables the statistical classification method to adapt to changing traffic patterns while maintaining accuracy by learning from past classification outcomes and refining its models over time.
Solution Approach 2:
The patent performs preliminary classification using IP/port numbers and statistical features before applying more complex analysis. This preliminary action filters out easily classifiable traffic, allowing the system to focus computational resources on ambiguous cases and improving overall accuracy while maintaining adaptability to traffic pattern changes.
4Reliability
If higher quality levels are applied to all data types, then service quality is improved, but network resource efficiency deteriorates
Solution Approach 1:
The patent implements local quality by applying different quality levels and processing intensities to different data types based on their specific requirements. High-priority traffic such as voice over IP receives enhanced processing and guaranteed bandwidth, while less critical traffic receives standard processing. This localized quality approach ensures that service quality is improved for critical applications without wasting network resources on non-critical traffic.
Data Source
AI summary
An approach for classification of a data traffic session over a data communications network is provided. A network device receives data packets of the data traffic session. The network device determines a plurality statistical factors that characterize the data traffic session based on the received data packets. The network device analyzes the determined statistical factors in relation to a plurality of classification variables, and classifies the data traffic session as being one of a plurality of session priority types based on the analysis, wherein each of a plurality of subsets of the classification variables delineates the plurality of session priority types over a continuum with respect to a corresponding one of the statistical factors.