Prevention of DDoS Attacks Targeting Financial Services using Supervised Machine Learning and Stacked LSTM

Distributed Denial of Service (DDoS) attacks are the most frequent and serious security threats to financial services. It is an attack-type that can be carried out fairly easily with the intention of making the resources unavailable to the users at the required time, but it is very difficult and costly to track and mitigate those attacks. In the recent past, many ISPs have adapted technologies to prevent DDoS at the network level, but they are not able to fully detect and mitigate many of the DDoS attacks as they originate from different types of devices and from several countries. There have also been several research strategies targeting mitigation of DDoS attacks in cloud computing environments, but none of them have tried to specifically target financial services transactions. Financial services are one of the most affected services among all the services in cloud computing environments. This research was conducted with the objective to mitigate DDoS attacks that target the financial services transactional endpoints of web applications, using stacked long short-term memory. This study mainly focused on the usage of deep neural networks to classify and mitigate DDoS attacks that are happening in the HTTP layer of a server, because deep learning neural networks have powerful performance when compared with other machine learning algorithms and intrusion detection techniques. This research study focused on three specific types of deep neural networks called long short-memory, gated recurrent unit, and simple artificial neural network. Keras and TensorFlow libraries were used to construct and train a stacked long short-term memory to discover the existence and DDoS types of attacks on a server. Moreover, during the research study, a high accuracy deep learning model was developed to predict and mitigate DDoS attacks in the application layer of transaction endpoints of financial services organizations. The main goal of the model produced in this research is to create a high accuracy and low false rate deep learning model for the mitigation of DDoS attacks targeting financial transactions on cloud computing environments.