Complex application identification and private network mining algorithm based on traffic-aware model in large-scale networks
| Autor: | Hui Zhu, Rongyu Tian |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Computer Networks and Communications
Computer science Process (computing) 020206 networking & telecommunications 02 engineering and technology Filter (signal processing) Flow network computer.software_genre Identification (information) Dimension (vector space) Multiple time dimensions 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Data mining computer Protocol (object-oriented programming) Software Private network |
| Zdroj: | Peer-to-Peer Networking and Applications. 12:1594-1605 |
| ISSN: | 1936-6450 1936-6442 |
| Popis: | Due to the lack of definition and analysis of complex applications in the field of protocol identification research, complex application identification is still attributed to traditional protocol identification, resulting in poor recognition performance. This paper has conducted in-depth research on complex application identification problems. Complex applications are first defined and distinguished from traditional protocols. At the same time, the communication process of complex applications is deeply analyzed, and the communication characteristics of complex applications are summarized. Then, based on the communication characteristics, a traffic-aware model describing the communication process of complex applications is proposed. The communication modes of complex applications are modeled from spatial dimension, time dimension and traffic dimension. Based on the traffic-aware model, spatial dimension awareness is used to filter network traffic. Finally, the traffic dimension is used to cluster the filtered multiple network flows into multiple network flow clusters and extract statistical features. The time dimension is used to construct the behavior state sequence of the complex application based on the statistical characteristics of the network flow cluster, and finally the behavior state is used. Sequences are used as identification features to effectively and accurately identify complex applications. The experimental results show that the accuracy of Skype recognition is improved from 25% to 80% after the original method is improved. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink