ITC-Net-blend-60: a comprehensive dataset for robust network traffic classification in diverse environments.

Area of Science:

• Computer Science
• Network Security
• Data Science

Background:

• Mobile application recognition in encrypted network traffic is crucial for network administration, security, and digital marketing.
• Existing datasets are limited as they originate from single network environments, hindering model evaluation for real-world robustness.
• Developing adaptable network traffic classifiers for dynamic settings remains a significant challenge.

Purpose of the Study:

• To address the limitations of current datasets by creating a more comprehensive and varied mobile network traffic dataset.
• To facilitate the development and evaluation of robust mobile application recognition models.
• To support research in network security and traffic analysis.

Main Methods:

• Collected traffic data from 60 popular Android applications across five distinct network scenarios.
• Varied scenarios by changing Internet service provider (ISP), geographic location, device, application version, and user.
• Captured traffic via real human interactions on physical devices, excluding background traffic and without requiring root access.

Main Results:

• The dataset contains over 48 million packets, 450,000 bidirectional flows, and 36 GB of data.
• Traffic was generated from 60 applications under diverse network conditions.
• Data collection method ensured practical applicability and avoided privileged access.

Conclusions:

• The newly created dataset offers a valuable resource for training and validating mobile application recognition models.
• Its diversity across network conditions enhances the evaluation of classifier robustness and real-world performance.
• This work contributes to advancing network security and intelligent network management through improved traffic analysis.