A Time Granular Analysis of Software Defined Wireless Mesh Based IoT (SDWM-IoT) Network Traffic Using Supervised Learnin
- PDF / 6,139,673 Bytes
- 27 Pages / 439.37 x 666.142 pts Page_size
- 56 Downloads / 205 Views
A Time Granular Analysis of Software Defined Wireless Mesh Based IoT (SDWM‑IoT) Network Traffic Using Supervised Learning Rohit Kumar1 · U. Venkanna1 · Vivek Tiwari1
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract The ceaseless increase in the number of the wireless Internet of Things (IoT) devices has resulted in the need of different traffic engineering techniques to manage the massive network traffic. Wireless Mesh Networks (WMNs) are an important constituent part of the wireless IoT networks, and are helpful to route the IoT networks’ traffic over long distances. The WMN devices are powerful in comparison to the IoT sensor devices, and are suitable to run the traffic engineering algorithms. To further improve the performance of the WMNs, Software Defined Networking can be used. Its unique features like global visibility, agility, etc., guarantee the optimal network management. As granularity plays an important role in data analysis and none of the existing works has discussed a time granularity based network data analysis, this work tries to offer a time granular analysis of Software Defined Wireless Mesh based IoT (SDWM-IoT) network’s traffic using supervised learning approaches. A time granular analysis helps to explore the functional traits of the data at the Coarse, Medium, and Fine granularity levels. This assists in divulging and understanding the hidden characteristics and behaviour of the SDWM-IoT network’s data based on varying time granularity, respectively. Some well known supervised learning algorithms are used to offer an in-depth analysis of the traffic, and to draw the relevant conclusions. Different variants of Decision Tree, Support Vector Machine and K-Nearest Neighbour (KNN) are used to analyze the traffic and achieve a reliable accuracy rate of more than 90%. Among all the variants, fine-KNN produces the best accuracy for most of the traffic classes with a rate of more than 98%. In addition to this, a tenfold cross-validation technique is also used to prevent the the chances of over-fitting. Keywords Time granularity · Internet of things (IoT) · Traffic engineering (TE) · Traffic classification (TC) · Software defined networking (SDN) · Wireless mesh networks (WMNs) · Supervised machine learning * Rohit Kumar [email protected] U. Venkanna [email protected] Vivek Tiwari [email protected] 1
Department of Computer Science and Engineering, DSPM-IIIT, Naya Raipur, India
13
Vol.:(0123456789)
R. Kumar et al.
1 Introduction The growing number of Internet of Things (IoT) devices have resulted in mammoth amount of data [1]. As most of the IoT devices are wireless in nature, the major part of this data is wireless in nature [2]. This data need to be analyzed properly to draw the meaningful heuristics that help to improve the system administration and performance. Thus, some Traffic Engineering (TE) approaches are needed to examine the IoT networks’ data and produce some critical conclusions [3]. But, the TE approaches involve heavy data processing
Data Loading...
