The Challenges in ML-Based Security for SDN

Machine learning (ML) is gaining popularity in the network security domain as more network-enabled devices get connected, as malicious activities become stealthier, and as new technologies like Software Defined Networking (SDN) emerge. From the application layer, ML-based SDN security models control the routing/switching of an entire Software Defined Network. Compromising the models is hackers' desirable goal. Previous works have been done on either adversarial machine learning without the context of secure networking environment or on the general vulnerabilities of SDNs without much consideration of the defending ML models. Through examination of the latest ML-based SDN security applications, a good look at ML/SDN specific vulnerabilities accompanied by a successful attack on StratosphereIPS, this paper makes a case for more secure developments of ML-based SDN security applications.