Machine Learning Based Intrusion Detection of DDoS Attack on IoT Devices

In an attempt to speed up productivity and also make life easier and comfortable for people, almost every sector of our life is continuously been connected to the internet. By so doing, data is quickly accessible to people by offering a centralized system where every device is connected and this is referred to as internet of things, IoT for short. Hence, the word smart is attached to things that are found in this centralized system. Smart hospitals, smart cars, smart homes are some examples. However, most of the devices that are connected to the internet are very poor when it comes to security and one of the dangerous security threats of IoT devices is Distributed Denial of Service attack simply DDoS attack. DDoS attack aims to prevent legitimate users from getting access to a targeted system service by exhausting the resources, bandwidth and so on. What makes DDoS attack very dangerous is the fact that there are so many free software over the internet that can even allow people with little to no computer skills to launch such an attack. Though, there are different mechanisms for the detection of DDoS attack but having an automated system that can learn the nature of the attack and instantly detect it is the reason why machine learning is used in this work. A supervised machine learning was employed in the training and analysis of the dataset. Decision tree, KNN and Naïve Bayes are the algorithms used to classify a benign traffic from a DDoS attack. About nineteen different features was carefully selected from CIC2019DDoS dataset. The DDoS attack types used for the experiment are UDP, DNS, SYN and NetBIOS. The results of the experiment indicate that Decision tree and KNN proved to be the most effective with an accuracy of 100% and 98% respectively. Naïve Bayes gave a very poor result with an accuracy of 29%.