A Deep Learning Model for Securing Software Defined Networks

Conventional computer networks have undergone a paradigm shift in terms of the advent of wireless pervasive networks such as IoT and fog networks. The ease of mobility and adaptive configuration enables significant ease of deployment and use of wireless networks over wired networks. However, the associated challenge remains the fact that wireless software defined networks (SDNs) are more prone to attacks from adversaries due to the absence of a secured communication medium. The SDN framework allows for a completely software based control plane of the network, which when coupled with stochastic computing can be leveraged to analyze network data. The analysis of data passing through an adversarial channel can be used for identifying potential attacks on the network. This paper presents a deep learning model for analyzing channel attributes to estimate potential adversarial activity and secure data transmission. The performance metrics of the system has been chosen as the error rate and sum secrecy rates. Comparing the performance of the proposed system with benchmark models indicates improved performance of the proposed system.