DDoS Detection using Deep Learning

Abstract

The network's infrastructure becomes more vulnerable to cyber-attacks as the number of services offered through the internet expands. The complexity of "Distributed Denial-of-Service (DDoS)" threats on the internet has recently increased, posing a challenge to typical protection systems. As a result, early identification and separation of network data is the most crucial part of protecting against DDoS threats. A "Long Short-Term Memory (LSTM)" based model is created in this study to identify DDoS threats on a sample of network traffic packets. LSTM is a deep learning technique that includes a feature selection and extraction algorithm. When trained, it updates itself; Even with a smaller number of data points, LSTM functions swiftly and correctly. Using the "CICDDoS2019 dataset" for training and testing, the suggested LSTM model can achieve an accuracy of up to 98 percent in the current work, and Deep learning exceeds machine learning on the CICDDoS2019 dataset.