Enhanced Convolutional Neural Networks for Adaptive IoT Botnet Attack Detection Using Reinforcement Learning

The proliferation of Internet of Things (IoT) devices has revolutionized various industries but has also exposed critical vulnerabilities in network security, making IoT networks prime targets for botnet attacks. This project addresses the challenge by developing a hybrid detection system that integrates Convolutional Neural Networks (CNN) and Reinforcement Learning (RL) for efficient and accurate botnet detection. Utilizing the N-BaIoT dataset, the system is designed to process large volumes of network traffic data, enabling the identification of anomalous behavior indicative of botnet activities. The architecture begins with data collection and preprocessing to ensure the input is normalized and structured for machine learning models. CNNs are employed for feature extraction, leveraging their ability to automatically learn complex patterns in IoT traffic data. This is followed by fully connected layers for classification, predicting whether a traffic instance is benign or botnet-related. The system achieves high accuracy through the hierarchical combination of convolutional layers and dense layers, ensuring robust detection capabilities across diverse traffic patterns. To enhance adaptability and decision-making, an RL agent is integrated into the pipeline. The RL agent employs a feedback mechanism, continuously learning and optimizing the model's performance based on real-time classifications.