Data Analysis for Predicting Attacks in Smart Systems Using AI Techniques

This paper presents an AI-driven framework to predict and mitigate DDoS attacks in smart cities using SoftwareDefined Networking (SDN), federated learning, and advanced machine learning algorithms. Parameter optimization significantly enhanced model performance. The Decision Tree algorithm, optimized via Research, achieved an accuracy of 0.9236 and an F1-Score of 0.929, outperforming Logistic Regression. Deep Learning models also improved with strategies like Early Stopping and ReduceLROnPlateau, with SimpleRNN showing the highest improvement, followed by GRU, LSTM, and CNN models. These results surpass previous research, demonstrating the effectiveness of the applied techniques and model preparation.