Loading...
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 52058 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Jalili, Rasool
- Abstract:
- Today, smartphones became a universal technology, and mobile networks speed and bandwidth have increased. Additionally, some of the messengers support voice call services that are easy to use and economical. Many of the VoIP servers provide encrypted VoIP service. On the other side, ISPs need to recognize the VoIP flows to enforce their policies. In some case, ISPs should block illegal VoIP calls or fix a specific charge for them. Also, for providing a better quality of service, it’s necessary to decrease the delay of VoIP packets. Most related works have concentrated on the methods of detecting VoIP traffic that just worked on a specific application. Some other works detect VoIP by using IP & port. In this research, we have proposed a framework for detecting encrypted VoIP flows. In the proposed framework, we don’t use IP, port, and payload. In this work, we have used a convolutional neural network to classify VoIP flows. We feed the first 1000 inter-packet time of flows to the convolution neural network. After that, the network will recognize whether the flow is VoIP or non-VoIP. We have achieved 0.84 precision, and 0.79 recall compare to 0.73 precision and 0.71 recall of previous studies
- Keywords:
- Traffic Classification ; Encrypted Traffic ; Deep Learning ; Convolutional Neural Network ; Voice Over Internet Protocol (VOIP) ; Detecting Voice
- محتواي کتاب
- view