Application Flow Clustering for New Network App Identification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing deep packet inspection (DPI) technologies suffer from low application identification rates due to the need for manual updates of traffic feature databases when new applications emerge, exacerbating congestion and packet loss in private networks like enterprise and campus networks.

Innovation Solution

Perform flow behavior feature analysis on a flow table to cluster services based on IP addresses and port identifiers, determining application types without relying on a traffic feature database, enabling direct identification of new applications through IP and port analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If deep packet inspection technology is used to identify applications by matching traffic features with a database, then application identification can be performed, but the application identification rate remains low due to manual database updates when new applications appear

Engineering Contradiction:
Improveapplication identification rateVSAvoidability to identify new applications
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system performs self-learning by automatically analyzing flow behavior features and clustering services to identify application types without manual intervention. The clustering module automatically updates application type definitions based on observed traffic patterns, enabling the system to adapt to new applications autonomously.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system incorporates feedback mechanisms where flow behavior analysis results feed back into the application type database. The clustering module uses observed traffic patterns to refine and update application type definitions, creating a continuous improvement loop that enhances identification accuracy over time.

Inventive Principle:
Principle #23Feedback

2Productivity

If a traffic feature database is maintained for DPI technology, then existing applications can be identified, but manual updates are required when new applications emerge, reducing efficiency

Engineering Contradiction:
Improveapplication identification efficiencyVSAvoidtime for manual database updates
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary analysis of flow behavior features and pre-clusters services into application types before new applications cause congestion issues. This proactive approach allows the system to prepare identification rules in advance, reducing reaction time when new applications appear.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The clustering module automatically maintains the application type database by analyzing new traffic patterns and creating updated application type definitions without human intervention, eliminating the time-consuming manual update process.

Inventive Principle:
Principle #25Self-service

3Reliability

If manual updates of traffic feature database are performed for new applications, then application identification accuracy can be maintained, but network congestion and packet loss are exacerbated

Engineering Contradiction:
Improveapplication identification accuracyVSAvoidnetwork congestion and packet loss
Core Design Contradiction:
ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The system continuously analyzes flow behavior features and updates application type definitions in real-time, maintaining uninterrupted identification accuracy. This continuous operation eliminates gaps in identification that would otherwise occur during manual update cycles, preventing traffic management decisions from being delayed.

Inventive Principle:
Principle #20Continuity of useful action

Solution Approach 2:

The automatic clustering mechanism continuously maintains the application type database without requiring manual intervention, ensuring that identification accuracy is maintained while the system actively manages network traffic to prevent congestion and packet loss.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4012980B1Application identification method and apparatus, and storage medium
Publication Date: 2026.03.18 HUAWEI TECH CO LTD
  • EP4012980B1 patent drawingFigure 1~2
  • EP4012980B1 patent drawingFigure 3~4
  • EP4012980B1 patent drawingFigure 5~6

AI summary

This application discloses a method, apparatus and storage medium for application identification, and belongs to the field of communications technologies. A network device performs flow behavior feature analysis according to a flow table, to obtain a plurality of services. Each service includes one IP address and one port identifier, and one application may usually include a group of services. Therefore, the network device clusters the plurality of services according to the flow table and a domain name table, to obtain a plurality of application types, where each application type includes a plurality of services, and each application type corresponds to one application. Further, the network device may determine a label corresponding to each of the plurality of application types, where the label is used to identify an application to which a data flow belongs. According to the method provided in this application, an application to which a data flow belongs can be identified based on a flow behavior feature, and a traffic feature database is not required. When a new application appears, the network device identifies the new application based on an IP address and a port of a server accessed by the new application. This improves an application identification rate.