An Adam-Optimised Deep Neural Net for Predicting Vulnerability Attack in Software Defined Networking

Umar Umar Abdullahi, Isa Modibbo Ismail, Babangida Isyaku

Keywords: Software Defined Networking, Intrusion Detection, Neural Network, Feature Selection


Software Defined Networking (SDN) has simplified network management through the separation of the control plane from the data plane. SDN controller decides where traffic flow is forwarded while the data plane, composed of network switches that forwards flow packets based on the decision made by the control plane. Forwarding logic is stored on a specialized switch memory of limited capacity known as Ternary Content Addressable Memory (TCAM). The centrality of control and the limitation of the flow table are the two most vulnerable aspects of SDN. The former is exploited by flooding the controller with malicious requests to overburden the controller and pave way for malicious attacks while the latter involves exploiting the limited flow table overflow. This paper proposes an Adam-optimised Deep Neural Network-based model for predicting these two vulnerability attacks in Software Defined Networks setting. This approach was tested on the NSL-KDD dataset, achieving an accuracy score of 93%. Experimental results also showed that this approach exhibited favorable performance on other metrics relative to some popular Machine Learning techniques. We conclude that this approach shows strong potential for Adam-optimised Deep Learners for SDN vulnerability attack mitigation.