To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Adversarial Attack on Network Traffic using Machine Learning for Software Defined Networks |
|
Author |
Muhammad Shahzad Haroon and Dr Husnain Mansoor |
| Citation |
Vol. 26 No. 3 pp. 37-46 |
|
Abstract |
New a day’s machine learning is employed in several different scenarios. One such scenario is an IDS (Intrusion detection system) which filters malicious traffic coming from a network. Machine learning algorithms are trained to classify network packets coming from different sources and classify them as normal or attack data. But Machine learning-based IDS are prone to adversarial examples that trick the classifier into believing that attack data is normal data. This poses a great risk to all those IDS that use Machine learning. We developed an adversarial dataset from the well knows UNSW NB15. This adversarial dataset was crafted using the Clever Hans library. This adversarial dataset is tested on Random forest OVR and decision tree OVR. The accuracy of classifiers has dropped when tested on adversarial data. The results show that adversarial examples pose a great threat to any application that is deployed using Machine learning algorithms |
|
Keywords |
SDN, Security, Machine learning, Openflow, dataset, Adversarial attack |
|
URL |
