Intrusion Detection System: The Use of Neural Network Packet Classification
Nery Ruiz, Bryan Tavera, Abdelshakour Abuzneid · 2020
Despite recent advances in cloud processing power and network connectivity to handle massive network traffic, networks are still vulnerable to Distributed Denial of Service (DDoS) attacks. With the recent proliferation of the Internet of Things (IoT), unsecured devices are fueling the ever-growing botnets, which allow creating larger malicious networks. Current mitigation techniques need to adapt to a new growing size of zero-day attacks to protect network services to consumers and block malicious connections. Deep learning enables machines to find the solution to many complex problems. This paper evaluates the performance of the Simple Neural Networks, Convolutional Neural Networks, and Recurrent Neural Networks in detecting DDoS attacks when trained with the CSE-CIC-IDS2018 Dataset. This research will discuss the presented datasets and the efficiency of the proposed networks. The trained data was obtained from a realistic dataset that holds different forms of intrinsic volume, protocol, and web-based attacks.