Paper Title
RESEARCH ON AUTOMATIC DETECTION OF NETWORK SECURITY THREATS USING DEEP LEARNING TECHNOLOGY

Abstract
This paper provides an in-depth look at the application of advanced machine learning models in Internet of Things (IoT) security, with particular emphasis on two key datasets: the widely used IoT-23 dataset and the specialized CRAFTED –Cooja RPL Attack Framework Test and Evaluation Empirical analysis of simulated attacks conducted by Dataset. This study develops multiple machine learning models to predict multiple IoT attack parameters and types. In the experiments, we evaluated a series of deep learning models, including long short-term memory network (LSTM) and residual network (ResNet), and compared the performance of these models with traditional models. Research results show that although deep learning models exhibit efficient learning capabilities when processing complex data, random forest and decision tree models exhibit superior performance in terms of real-time processing speed and accuracy, making them the best choices. In addition, this research also uses advanced feature engineering and hyperparameter optimization technology to significantly improve the model accuracy and provide valuable technical support and practical guidance for the detection of real-time network security threats in the Internet of Things environment. This research not only promotes the development of IoT security technology, but also provides a solid foundation for future research in related fields. Keywords - Internet of Things (IoT), Anomaly Detection, Decision Tree, Deep Learning, Network Security, LSTM (Long Short-Term Memory), ResNet (Residual Network), IoT-23 Dataset.