Machine learning techniques for traffic identification and classification in SDWSN: A survey

Abstract

Software defined network (SDN) is a paradigm developed achieve great flexibility and cope with the limitations of traditional networks architecture such as the wireless sensor networks (WSNs). Introducing SDN in WSN leads to SDWSN. However, due to the challenges that are inherent in SDN and WSN, SDWSN is faced with number of challenges such network and Internet traffic classification (TC). Several solutions have been offered such as machine learning (ML) technique but there are several challenges that still exist which need attention. Therefore, this paper present a review on the approaches of TC in SDWSN using ML and their challenges. The objective is to identify existing approaches and the challenges in order to provide ways to enhance them. We performed review of the existing works on TC in the literature based on the aspect of enterprises network, SDN and WSN has been done as well as findings reported. Our findings shows that the approaches to TC using ML were based on supervised or unsupervised learning. Moreover, TC is faced with challenges which include energy efficiency, shareable test data and design. Thus, ML technique to TC in SDWSN is still at its early stage and need to improve in order to accurately classify traffics that normal or abnormal.