A MACHINE LEARNING ALGORITHM WITH SELF-UPDATE PARAMETER CALIBRATION TO IMPROVE INTRUSION DETECTION OF DDOS IN COMMUNICATION NETWORKS

Abstract

Currently DDoS attack has become one of the most common network attacks worldwide. This is largely due to the fact that we live in the age of the Internet of Things, with the rapid development of computer and communication technology evolving into big, complex and distributed systems that are exposed to several kinds of attacks in addition to new threats. In order to detect intruders in an efficient and timely manner, a real time detection mechanism, proficient in dealing with a variety of forms of attacks is highly important. However, due to the uniformity and evolution of DDoS attack modes and the variable size of attack traffic, there has not yet been a detection method with satisfactory detection accuracy at present and considerable effort made by both the scientific research and industry for several years to mitigate DDoS detection potential DDoS target indicate that DDoS attacks have not been fully addressed. This study therefore aimed at developing a machine learning a Machine learning algorithm with self-update parameter calibration to improve intrusion detection of DDoS in communication networks, in two steps: Feature extraction and model detection that is, we extract DDoS attack traffic characteristics with large proportion and compare the data packages according to the protocol in the Feature extraction stage whereas in the model detection stage, the features that were extracted are used as the input features in machine learning after which the Random Forest algorithm used to train the developed detection model. Finally, the model was validated by three metrics (accuracy, false negative rate and false positive rate). The results show that the DDoS attack detection method based on machine learning proposed in this study has a good detection rate and accuracy compared to the current popular DDoS attack detection methods. The developed model achieved accuracy of 96% over a real-time dataset.