Identifying optimal feature pairs and detecting anomalies using machine learning on software-defined networks

One of the biggest problems of today's internet technologies is cyber attacks. Although information technologies have made great progress in recent years, traditional networking principles have hardly changed. In our study, efficient monitoring of network traffic was achieved thanks to its centralized control and programmability using Software-Defined Network. Any fully-featured machine learning detection system turns into inefficient and heavy intrusion detection systems. Therefore, light, accurate and high-performance intrusion detection systems are very important instead of inefficient and heavy intrusion detection systems. For this purpose, the 10 most popular machine learning algorithms and BoT-IoT (2018) dataset were selected. The twelve best features recommended by the developers of this dataset are used in this study. Similarly, 660 feature-pair-based lightweight intrusion detection systems were developed by training the 10 machine learning algorithms via each feature pair out of the 66 feature pairs. Moreover, the 10 intrusion detection systems trained with 12 best features and the 660 intrusion detection systems trained via 66 feature pairs were compared to each other based on the machine learning algorithmic groups. While the results obtained with feature pairs are over 95%, it also provides 20%-30% savings in terms of unit cost. Keywords: Software-Defined Networks, Cyber Security, Bot-Iot Dataset, Machine Learning