A New Distributed Denial-of-Service (DDoS) Attacks Detection System Combining Multistage Auto-Encoders with Radial Basis Function (RBF)

Distributed denials of service (DDOS) attack have strong impact on the cyber world. As far as cyber-attack is concerned that it halts the normal functioning of the organization by Internet protocol (IP) spoofing, bandwidth overflow, consuming memory resources and causes a huge loss. There has been a lot of related work which focused on analysing the pattern of the DDOS attacks to protect users from them. A User datagram protocol (UDP) flood is a network flood and still one of the most crucial network floods today. This study presents new method to detect DDOS attacks by using multistage auto-encoders based on Radial Basis Function (RBF). The input data which represented the DDoS features are first analyzed by using autoencoders and the number of auto-encoders depended on the data nonlinearity and dimension. The output of the first auto-encoder wired to the second auto-encoder etc. The aim of these auto-encoders are to extracted features that have ability to presented the best classification results and to speed up the processing time by reducing the dimension of features. In the last stage, the Radial Basis Function (RBF) trained in supervised method to classify the features into two labels there is attack or not. The obtained results compared with wellknown studies presented in this field.