“Using deep learning to detect DDoS attacks in 5G networks.”

Abstract

The conventional ideas of detecting such attacks are fast becoming obsolete because of the ever-evolving dynamics of 5G network traffic. This paper aims to investigate the most effective use of deep learning to identify DDoS threats in 5G networks. The study with CICIDS2017 and UNSW-NB15 datasets applies both CNN and LSTM models on the network traffic classification. These statistics show that the approaches identify DDoS attacks with great precision and often (98% for CICIDS2017 and 94% for UNSW-NB15), and remember them well (96% and 92% receptively).

The paper shows that the model successfully learns traffic patterns characteristic of DDoS attack and surpassing the performance of traditional approaches. But issues like the noisy database and model versatility were raised, meaning that there is plenty of scope to make the training data more diverse. As inferred from the results, it is possible to greatly improve the protection of the 5G networks from DDoS attacks with the help of the integration of deep learning into network security systems. As the evolution continues, more attention should be paid on enhancing the model stability and expanding the training datasets for attacking more diverse scenarios that can become a supreme benefit of using deep learning in securing 5G networks.