Mitigating Distributed Denial of Service (DDoS) Attacks in Cloud Networks Using Neural Networks
Jagendra Pratap Singh, Jabir Ali, Preeti Sharma, Vinish Kumar, Meenakshi Sharma, Ramendra Kumar Singh · 2024
The research aims to prove the effectiveness of three neural network models the convolutional neural network (CNN) and Long Short-Term Memory (LSTM) Hybrid, Variational Autoencoders, and Bidirectional LSTM in neutralizing DDoS attacks on online shopping websites. The models are ranked using critical performance metrics relevant to cybersecurity. Firstly, the CNN-LSTM Hybrid has demonstrated a very high True positive rate (TPR), with a reading of 0.95, as well as a low False positive rate (FPR) of 0.03, resulting in a precision of 0.93 and a recall of 0.95. The readings show that the model is accurate in distinguishing between regular traffic and DDoS attacks, having an overall Fl-Score of 0.94. On the anomaly detection-based Variational Autoencoders (VAEs), the TPR was 0.92, the FPR was 0.05, the precision was 0.90, and the recall 0.92. The bidirectional LSTM model returned a TPR of 0.94 and an FPR of 0.04, and both precision and recall were also brought to 0.92 with an Fl-score of 0.93. These models improve online shopping website security from cyberattacks by preventing data availability and consistency. As DDoS attack discovery is a major step for continuing institutions, the designs serve administrators and greatly protect online buyers.