TY - GEN
T1 - A modular two-layer system for accurate and fast traffic classification
AU - Hajikarami, Fateme
AU - Berenjkoub, Mehdi
AU - Manshaei, Mohammad Hossein
N1 - Publisher Copyright:
© 2014 IEEE.
PY - 2014/12/18
Y1 - 2014/12/18
N2 - Traffic classification is the basic block of many network management and control operations such as allocating appropriate levels of quality of service to different applications, filtering and intrusion detection. Due to importance of accurately classifying large networks traffic, we have proposed a lightweight two-layer system, which is perfect for today's high-speed links. Considering the trade-off among accuracy and speed, our system is based on a modular architecture and combination of some expert's opinions. Moreover we extract the best features for each application group in order to achieve an accurate system. Experimental results indicates that our proposed system has accuracy comparable to more sophisticated system and its learning cost is 10 times less than these systems. In addition we can customize our proposed system according to the needs and goals of network administrator. Our system is robust and adaptable to network changes such as the appearance of new applications. In such a case, our system labels new flow as 'unknown' instead of misclassifying. So the administrator will notice and take necessary actions to improve the classification system.
AB - Traffic classification is the basic block of many network management and control operations such as allocating appropriate levels of quality of service to different applications, filtering and intrusion detection. Due to importance of accurately classifying large networks traffic, we have proposed a lightweight two-layer system, which is perfect for today's high-speed links. Considering the trade-off among accuracy and speed, our system is based on a modular architecture and combination of some expert's opinions. Moreover we extract the best features for each application group in order to achieve an accurate system. Experimental results indicates that our proposed system has accuracy comparable to more sophisticated system and its learning cost is 10 times less than these systems. In addition we can customize our proposed system according to the needs and goals of network administrator. Our system is robust and adaptable to network changes such as the appearance of new applications. In such a case, our system labels new flow as 'unknown' instead of misclassifying. So the administrator will notice and take necessary actions to improve the classification system.
KW - Combining methods
KW - Machine learning
KW - Modular architectur
KW - Statistical features
KW - Traffic classification
UR - http://www.scopus.com/inward/record.url?scp=84921060032&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84921060032&partnerID=8YFLogxK
U2 - 10.1109/ISCISC.2014.6994039
DO - 10.1109/ISCISC.2014.6994039
M3 - Conference contribution
AN - SCOPUS:84921060032
T3 - 2014 11th International ISC Conference on Information Security and Cryptology, ISCISC 2014
SP - 149
EP - 154
BT - 2014 11th International ISC Conference on Information Security and Cryptology, ISCISC 2014
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2014 11th International ISC Conference on Information Security and Cryptology, ISCISC 2014
Y2 - 3 September 2014 through 4 September 2014
ER -