Detection of DDOS Attacks in IIoT Case Using Machine Learning Algorithms

Internet of Things has grown very large and fast over the past few years. One proof of the development of IoT is the use of Industrial Internet of Things in the industrial field. IIoT can help improve the production process and the resulting profits. However, IIoT also has a weakness on the security side. IIoT is very vulnerable to attacks, especially Distributed Denial of Service (DDoS) attacks. DDoS can inhibit data transmission and thus slow down the performance of IIoT. This can cause huge losses for the industry itself. In order to prevent these attacks, one way is to detect them using machine learning. Therefore, research is conducted to test the detection of DDoS attacks on IIoT using machine learning algorithms. The research will use Random Forest and Naive Bayes algorithms as detection models and tested using the Edge-IIoTset dataset. These two algorithms are chosen because both have been proven to detect DDoS attacks on various types of networks such as web networks, cloud-servers, IoT networks and others. In the testing process, two scenarios will be carried out, namely testing a different number of features and testing using a different number of data split. The resulting performance level in testing with random forest reaches 100
^%
for each matrix and test scenario. Meanwhile, the performance level with naive bayes is different for each test scenario. The best performance level of naive bayes is with a total of 20 features that can achieve 78.56% for accuracy, 79.85% for precision, 90.61 % for recall, and 84.89% for f-1 score. Based on these results, the best method that can be used to detect DDoS attacks on IIoT is the random forest method.