A Recurrent Neural Network Framework for Effective DDoS Attack Detection in Cloud Computing
Nidhi Soni, Lakshmi Chandrakanth Kasireddy, S. Theetchenya, Chetan Sinhgadiya, Sheo Kumar, T. S. Arulananth · 2025
A Distributed Denial of Service (DDoS) assault is an intrusion into a cloud computing environment that secretly inserts malicious data packets into client-side internet traffic, causing a catastrophic impact on end users. The detection and prevention of infiltration is a tough issue that will influence the operation of the whole architecture, making it a critical concern in cloud computing. There have been a lot of cyber-security checks done to make sure the server is safe from hackers and other harm. Traditional cyber-security measures were ineffective in protecting the system from several external illegal traffics. Improving IoT architecture with an IDS is crucial. The goal of doing thorough literature studies is to study different machine learning approaches, neuronal system replicas, and optimization algorithms. The goal is to discover any gaps or flaws and then design machine learning algorithms that can reliably and effectively detect intrusions. In comparison to its rivals, the proposed LSTM model achieved superior accurate classification and faster convergence. An over-fitting problem may plague deep learning models, notwithstanding their usefulness when dealing with large datasets. It is overcome by fixing layers that are hidden and hidden neurons using a trial-and-error process. At 98.73% accuracy, the suggested technique Long Short-Term Model (LSTM) more than doubles the performance of the competing models. Compare this to Random Forest (RF), which delivers intermediate accuracy, Support Vector Machine (SVM), with the least accuracy, and Naive Bayes (NB), which offers great accuracy. By and large, the suggested approach outperforms the competition when it comes to accuracy.