Type: Process Essays
Sample donated: Jo George
Last updated: July 26, 2019
Mr THANIGAVEL M_1,Ms .JAYASREE K2_,Ms.KOWSALYA.S_3,Ms.PRIYADHARSCINI S_4 1AssistantProfessor, Dept. of Information Technology, Prathyusha Engineering College.
2 3 4 Students, Dept. ofInformation technology, Prathyusha Engineering College. ABSTRACT: Traffic classification is acentral topic in the field of computer science. The Classification and theanalysis of the network traffic is useful to avoid the traffic congestion whiletransferring the data. The traffic classification refers to categorizing thetraffic according to its various application type and also helps in managingthe overall performance of a network.
. The Traffic flow analysis is anessential piece of knowledge for engineering a network. However, with the rapidevolution of the internet applications the effectiveness of the traditionalmethods like port based, payload based. The Machine learning algorithms hasachieved high accuracy and best results.
The use of Hello Packet for theclassification makes easier in analysing the real-time network which addsaccuracy to the existing system. The accurate classification is achieved usingthe hello packet which classifies the network in a effective way and transfersthe data through the easier and the shortest path.Keywords: Traffic Classification, Hello Packet, TrafficFlow Analysis.
I.INTRODUCTION Network Trafficclassification is an important topic nowadays in the field of communication andcomputer science.Thisis useful for the Internet Service Providers(ISPs) to manage the network performance.Traffic classification is the first step in identifying and classifying theunknown network classes. TheNetwork traffic classification plays a vital role in network security andmanagement such as the Intrusion Detection and Quality of Service ( QoS) Thereare several techniques have been proposed classifying and analysing the networktraffic which includes the traditional techniques like port based methodpayload based techniques these techniques .Portbased technique is a great technique for network classification . This technique failed due to increase of peerto peer applications.
The Payload based Technique which is alsocalled Deep Packet Inspection (DPI) is effectivein classification but it cannot be applied to encrypted data networkapplications as numerous data network applications use encrypted techniques toprotect the data from detection . The DPI technique failed due to use ofencrypted flow of applications . The researchers then proposed MachineLearning Technique to classify theinternet traffic as well as to know the type of applications flow in thenetwork. II.PROBLEM STATEMENT The evolution in the internetapplications has led to different methods of classifying the network traffic. Theport based, payload based techniques has efficient classification but comparativelyhas low accuracy. These techniques failedfor the encrypted data network applications. The port based technique faileddue to dynamic port numbers.
Dynamic port number means unregistered number withthe (Internet Assign Number Authority). The classification results in a betterway but it doesn’t compare the results of those algorithms. The port basedtechnique has failed due to increase of peer to peer communication. III.
EXISTING SYSTEM The network classification hasseveral traditional techniques to classify the network traffic such as thepayload method and port-based method which does not support for encrypted datanetwork applications as numerous network applications use encrypted network toprotect data from detection. Port based techniques failed due to increase ofpeer to peer applications. The port based and the payload based techniques haveachieved in classifying the network effectively but has failed when applied toencrypted data applications. IV.
PROPOSED SYSTEM TIE recommends a unified representation ofclassification results.It defines IDs for application classes and associatesthem with group classes.The comparison of traffic classifiers which have application-levelprotocol.It uses the Hello Packet to compute the distance from the source todestination. The hello packet establishes and confirms networkrelationship.It distinguishes the traffic as per its constraints.TheTIE is also used to compute the following Duration of a video or audio stream Voiceor video quality of experience Countsof the number of events Trackingof “top” items (e.
g., most frequently requested URLs, most popular videoproviders, etc.) Summations (e.g.
, adding up a number ofevents). IV.OBJECTIVEThemain aim of the project is to transfer the data without traffic congestion. Thisanalyses the traffic and classifies the network traffic and transfers the dataover the shortest path. The shortest path is computed using the MachineLearning algorithms. TheMachine Learning algorithm is used as it achieves high accuracy.
The HelloPacket is used to easily classify the network traffic and transfers the datavia the free and the shortest Path. V.METHODOLOGY Hello Packet implementation alongwith the machine learning algorithms provide higher accuracy compared toprevious classification techniques.HelloPacket is a special packet that is sent from a router which is used toestablish and confirms the network adjacency relationships.TheHello Packet classifies the network traffic and transfers the data via theshortest path. The shortest path is computed using the Machine Learning algorithms.The path is classified as traffic free network and network with traffic and itthen chooses the easier and the shortest path.
ARCHITECTURE DIAGRAM TheSender analyses and captures the available ip on the network from which thesender ip is been selected. The File to be sent is been attached and the fileis sent.The analyser is that finds the available nodes and lists all therouters.The free router from that is chosen to transfer the file.Theclassifier classifies the path based on the implemented machine learningalgorithms and finds the shortest path from that of the available routers.
Thisautomatically identifies the path or it can be done manually by the sender bywhich the path can be selected by the sender itself. The file that istransferred is received at the receiver’s end.` VI.
NETWORK TRAFFIC CLASSIFICATION MODEL A.Capturesthe IP This is the first and theforemost step, it captures the ip address of the sender. The File to betransmitted is selected. The content of the file is obtained it also lists thepacket length and size. The packet filter displays the size of the packet to betransmitted then the file is sent.
B. Traffic Classification The system starts analysingthe network traffic by which it identifies the ways to send the file. Itdistinguishes the free network from that of the path with traffic and makes thecommunication easier. C. Identifying the Path The Routers are listed which are used to transferthe file selects the path based on the machine learning algorithms .It findsthe shortest path and the free path among the available routers and displaysthose routers.
D.File TransferAfterthe process of finding the path the data that is selected is been sent throughthe free router and the file is been transferred to the receiver. After the process of findingthe path the data that is selected is been sent through the free router and thefile is been transferred to the receiver. The Receiver then receivesthe transferred file from the sender this also lists the contents of the file.. VII.RESULTSAND OBSERVATIONTheImplementation of Algorithms used hereachieve high accuracy.
There are four algorithms used to find the shortest pathon the network. They are C45, SVM, Bayes Net, Naive Bayes algorithms. Thecomparison of the accuracy is shown below. The file transfer is made easier.The use of the Hello packet is been analysed. Theways of classification and the comparison of the algorithms along with thehello packet has achieved high results of accuracy. accuracyis shown below.
The file transfer is made easier. The use of the Hello packetis been analysed. The ways of classification and the comparison of thealgorithms along with the hello packet has achieved high results of accuracy Classifiers Accuracy Time T in seconds C45 78.9189 0 Bayes Net 68.
1081 0.1 SVM 74.0541 0.3 Naïve Bayes 71.8919 0.1 Theefficiency of these four algorithms are analysed and the results are beingcompared and listed below in the form ofgraph.
VII.CONCLUSION In this paper, we discuss Network trafficclassification techniques and discuss How new researchers or new networkoperators will apply the network traffic classification technique using machinelearning algorithm to classify unknown applications and manage performance ofnetwork. And then we perform comparative analysis of four machine learningclassifiers. Firstly, we demonstrated Network Traffic Classification Techniques(Port Based, Pay Load Based and Machine Learning Based technique) and theirlimitation. The Hello Packet concept has been implemented which is used foreasier data transfer which adds additional advantage to the existing system.The Classification is done in a better way and achieves better accuracy. REFERENCE TABLE REFERENCES1. Hamza Awad Hamza Ibrahim, Omer Radhi AqeelAlZuobi, Marwan A.
Al-Namari,Gaafer MohamedAli, Ali Ahmed Alfaki Abdalla, InternetTraffic Classification using Machine Language Approach: Datasets validationissues IEEE 2016.2.Muhammad Shafiq, Xiangzhan Yu, Asif Ali Laghari , Lu Yao, Nabin Kumar Karn,Foudil Abdessmia, Network Traffic Classification techniques and comparativeanalysis using Machine Learning algorithms IEEE 2016.3.Walter deDonato, Antonio Pescape, Alberto Dainotti, Traffic IdentificationEngine: An Open Platform for Traffic Classification IEEE 2014.4.P.
Raj Kumar , P.Prasanna,Traffic Classification By Using :TIE(TrafficClassification Engine, International Journal of Engineering and ComputerScience(IJECS) 2015.5.Mohammad Reza Parsaei, Mohammad Javad Sobouti, Seyed Raouf Khayami, RezaJavidan, Network Traffic Classification using Machine Learning Techniques overSoftware Defined Networks International Journal of Advanced Computer Scienceand Applications (IJRSCA)2017.6.HE HaiTao, LUO XiaoNan, MA Fei Teng, CHE ChunHui & WANG JianMin,Network TrafficClassication based on Ensemble Learning and co-training,Springer 2009.7.Michael Finsterbusch, Chris Ritcher Eduardo Rocha,Jean-Alexander Muller,KlausHanssgen, A Survey of Payload-Based Traffic Classification Approaches.
Moore and K. Papaginnaki,Toward theaccurate identification of network applications in PAM 2005.9.A. Dainotti, W. De Donato, A.
Pescape, and P. Salvo Rossi, “Classification ofnetwork traffic via packet-level hidden markov models. IEEE GlobalTelecommunications Conference, 2008.
10.Ms. Zeba Atique Shaik,Prof.Dr.
D.G. Harkut, An Overview of Network TrafficClassification Methods International Journal on Recent and Innovation Trends inComputing and Communication(IJRITCC)2015.