Detection of LR-DDoS Attack Based on Hybrid Neural Networks CNN-LSTM and CNN-Autoencoder

Modern technologies improve each day. Accordingly, increasingly demanding, and modern services and applications require large, robust, and reliable network infrastructures. With the growth of medium and large network infrastructures such as Software Defined Networks, Data Center networks, and Internet of Things networks, we also see an increase in threats to them. One such modern threat is the Low-Rate Distributed Denial of Service (LR-DDoS) attack. LR-DDoS attack discreetly floods a target system with a minimal amount of malicious traffic to interrupt services, all while trying to avoid prompt detection. In this paper, we have presented modern hybrid neural network algorithms Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM) and Convolutional Neural Network - autoencoder (CNN-autoencoder), that can help in detecting these types of threats, as well as their capabilities and usage methods.