Presentation | 2015-09-04 [Encouragement Talk] Identification of Mobile Applications via In-Network Machine Learning Using N-gram for Application-Specific Traffic Control Takamitsu Iwai, Akihiro Nakao, |
---|---|
PDF Download Page | ![]() |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Identifying the application transmitting a given flow of network traffic is beneficial for network management, especially for achieving application specific QoS, enabling malware detection, and executing network functions such as content caching only for a particular application. Although typical methods for application identification include port scanning and pattern recognition using application signature, they suffer from various problems, e.g., for the former, ephemeral port usage and dynamic port allocation hinder accurateapplication identification, and for the latter, it is costly to collect application signatures, especially from encrypted traffic. The existing research for application identification using machine learning have shortcomings such as a limited scope of identifiable applications, inability to deal with real-time traffic, and few efforts have been put to fine-grained application identification, e.g., at the level of application identifiers such as YouTube and Chrome. We have proposed a real-time identification method using reliableand on-line training data collection performed by adding the application identifier at the end of the SYN packet. Our existing method identifies 80% mobile applications accurately without Deep Packet Inspection (DPI). In this paper, we propose a new method that improves inference accuracy using DPI even applicable to encrypted traffic. We evaluate our method in real MVNO traffic and show our method identifies at maximum 92% applications in the trafficaccurately using 2-gram features of packet payloads. We also improve the inference accuracy from 82% without DPI to 90 % with DPIwhen learning period is limited to 5 days. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | application identification / machine learning / SDN / NFV / MVNO / mobile edge computing |
Paper # | NS2015-78 |
Date of Issue | 2015-08-27 (NS) |
Conference Information | |
Committee | IN / NS / CS |
---|---|
Conference Date | 2015/9/3(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Iwate-ken Kokaido |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Post IP networking, Next Generation Network (NGN)/New Generation Network (NWGN), Contingency Plan/BCP, Network Coding/Network Algorithms, Session Management (SIP/IMS), Internetworking/Standardization, Network configuration, etc. |
Chair | Hidetsugu Kobayashi(NTT) / Atsushi Hiramatsu(NTT-AT) / Toshinori Tsuboi(Tokyo Univ. of Tech.) |
Vice Chair | Katsunori Yamaoka(Tokyo Inst. of Tech.) / Hideki Tode(Osaka Pref. Univ.) / Tetsuya Yokotani(Kanazawa Inst. of Tech.) |
Secretary | Katsunori Yamaoka(NTT) / Hideki Tode(KDDI) / Tetsuya Yokotani(Univ. of Fukui) |
Assistant | Yuichi Sudo(NTT) / Kunitake Kaneko(Keio Univ.) / Shohei Kamamura(NTT) |
Paper Information | |
Registration To | Technical Committee on Information Networks / Technical Committee on Network Systems / Technical Committee on Communication Systems |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | [Encouragement Talk] Identification of Mobile Applications via In-Network Machine Learning Using N-gram for Application-Specific Traffic Control |
Sub Title (in English) | |
Keyword(1) | application identification |
Keyword(2) | machine learning |
Keyword(3) | SDN |
Keyword(4) | NFV |
Keyword(5) | MVNO |
Keyword(6) | mobile edge computing |
1st Author's Name | Takamitsu Iwai |
1st Author's Affiliation | University of Tokyo(UTokyo) |
2nd Author's Name | Akihiro Nakao |
2nd Author's Affiliation | University of Tokyo(UTokyo) |
Date | 2015-09-04 |
Paper # | NS2015-78 |
Volume (vol) | vol.115 |
Number (no) | NS-209 |
Page | pp.pp.41-46(NS), |
#Pages | 6 |
Date of Issue | 2015-08-27 (NS) |