Video Stream Recognition Using Bitstream Shape for Mobile Network QoE.

Area of Science:

• Computer Science
• Telecommunications Engineering
• Network Security

Background:

• Mobile network operators face challenges in identifying video streaming services due to increasing encrypted internet traffic.
• Accurate service identification is crucial for quality of service (QoS) management, user experience enhancement, and implementing network policies like throttling or differentiated pricing.

Purpose of the Study:

• To develop and evaluate a novel method for recognizing video streams based solely on bitstream shape within cellular network communication channels.
• To address the limitations posed by encrypted traffic in traditional network traffic analysis.

Main Methods:

• A convolutional neural network (CNN) was employed for classifying network traffic bitstreams.
• The CNN model was trained using a custom dataset of download and upload bitstreams collected from real-world mobile network traffic.

Main Results:

• The proposed method demonstrated high accuracy, exceeding 90%, in identifying video streams from mobile network data.
• The approach effectively distinguishes video traffic even when the content is encrypted.

Conclusions:

• The bitstream shape analysis method using CNNs is a viable and accurate technique for recognizing video streams on cellular networks.
• This method offers a practical solution for mobile network operators to manage traffic and enhance user experience in the era of encrypted data.