Top-Performing Unifying Architecture for Network Intrusion Detection in SDN Using Fully Convolutional Network

Internet's omnipresent adoption over the course of many years leading up to the current era along with Wireless Sensor Networks' (WSN) limitations has only increased the security threats to these high-speed networks. As a result, it is critical to deploy Intrusion Detection Systems (IDS) to monitor malicious activities in the network framework. Software Defined Network (SDN) has served its purpose to a certain extent by enabling centralized control and management of WSN and providing ways to implement IDSs more efficiently. However, like any emerging technology, SDN also has its share of vulnerabilities as there exists a gap for contemporary architectures, which can exploit global network assets and balance network resource consumption while providing programmability in terms of security of the network. To tackle this problem, a model has been developed that uses a fitting feature selection method to cut down on the count of features extracted and minimize the data volume relayed over the routes. Conclusions drawn from the scheme's trials demonstrate the proposed IDS technology's potency and effectiveness. The openly accessible University of Nevada - Reno Intrusion Detection Dataset (UNR-IDD) was used to evaluate the intrusion detection algorithm involving the Fully Convolutional Network (FCN) architecture and its effect using a statistical technique in detecting different types of attacks.