Abstract:

In our time, billions of devices and applications based on the Internet of Things are interconnected at a high speed and are expanding continuously day by day until it has become acquiring every aspect of our daily life, with the aim of making human life more intelligent and more advanced, but due to the simple structure of this technology Relying mainly on the Internet, it has become easy for hackers to access it from all sides to pose serious threats to users' data and privacy. The Internet of Things is now facing important security challenges, which require a lot of concerns on the part of researchers and inventors. Machine learning algorithms help design "Intrusion detection models" that aim to detect attacks and classify network traffic into abnormal behaviors or normal traffic for devices and smart networks. In this paper we present a comparison paper according to the performance of Garrett ranking technique among  compatible different ML algorithms for NSL-KDD dataset, with 41 features and about 94,000 instances in training dataset with 48,000 instances as a test dataset, using WEKA tool (open source knowledge analysis software). It found that the Rotation Forest algorithm ranked first as the best performance in detecting attacks and anomalies.