An Autonomous System for Predicting DDoS Attacks on Local Area Networks and the Internet

Distributed denial of service (DDoS) attacks continuously evolve, causing losses and increasing service costs. The high volume and fast scaling make it difficult to defend from them. DDoS attack detection is not sufficient to protect the services from attacks. Thus, it is necessary to design prediction strategies to confront these attacks. When it is possible to identify the preparation for attacks, the time to combat them increases. This paper proposes a self-adaptable system to identify DDoS attack preparation and predict them. The system automatically determines the most appropriate neural network architecture for predicting attacks in different scenarios. The system predicts a DDoS attack with an accuracy of 97.89%, higher than the literature, and the prediction occurs 29 minutes before it starts.