Efficient UDP Flood Detection in SDN Using Feature Optimization and Deep Learning

ABSTRACT SDN (Software Defined Networks) utilizes cutting-edge dynamic technology to promptly and precisely track and manage the flow of data across a network. Unlike conventional networks, SDNs' packet traffic can be readily predicted and you have more options for controlling the network quickly. In this paper, SDN is used to detect UDP flood attack in the network level. It is simpler to monitor data flow from networks, analyze the network, and identify many forms of malware, DoS, and DDoS. To properly categorize various kinds of attacks based on their attributes, Deep Learning approaches have been used. In order to identify UDP attacks, several methods have been suggested. However, there has been hardly any investigation on SDN. With the Software defining network Layer attack detection framework in mind, we present a UDP attack detection solution for SDN. The Error Scalar maxy normalization is used as a preprocessing step and once those traits are identified, the Iterative random wrapper Boruta technique is used to extract the features that are relevant to detect the attack. Finally, we use a framework for identifying attacks based on CICDDOS2019 dataset for attack detection at the software-defined network layer. Our solution is tested in a software-defined-networking (SDN) setting built in Python and the TensorflowKeras application programming interface and the findings show an accuracy of 99.8% in predicting the attack in the networks.