Uses of Artificial Intelligence in Cyber Security to Mitigate DDOS

Abstract

The Distributed Denial of Service (DDoS) attacks have a negative impact on service availability and service providers, as well as on the economy. They lead to difficulty in the provision of online services because most or all the service providers use resources, and the bandwidth becomes completely depleted due to the lack of resources to perform the basic functions required in service delivery platforms. Because this kind of attack is unconventional, the detection process demands more complex methods and algorithms. Based on several algorithms such as Decision Tree, Naïve Bayes, K-Nearest Neighbors, and Support Vector Machine algorithms, in this study we supplied an ideal model for the detection of the attack. We used a training dataset that had been built from the known and suspicious traffic dataset for learning using different classification algorithms to make a prediction model to be utilized as a reference for the unknown traffic classification. To make sure we choose the best prediction model, we performed a validation for the prediction model by applying the prediction model to a test dataset, generating a confusion matrix, and utilizing it to obtain the maximum prediction probability and minimum false alarm rate. By using the prediction model, we can detect the attack traffic, get the necessary data about attacking botnet and forward this information to a firewall to block suspicious traffic and stop the attack.