As the Internet of Things (IoT) technology develops rapidly, the interconnection of various devices and sensors has brought great convenience to people's lives and work. However, at the same time, IoT systems also face security threats and intrusion risks. Therefore, designing an efficient and reliable IoT intrusion detection model has become an important issue that urgently needs to be addressed today. The main function of the perception layer of the IoT is to achieve intelligent perception of information acquisition, capture, and target recognition, and generate massive amounts of new data. In the IoT, the transmission and transmission of information are mainly completed through the network layer. At the application level, it can achieve fields such as smart grids, monitoring services, smart homes, environmental monitoring, public safety, green agriculture, and industrial monitoring. Intelligent processing and analysis of the large amount of data generated by the perception layer can effectively manage the risk of IoT intrusion. This article explores an IoT intrusion detection model based on convolutional neural network (CNN) algorithm and random forest (RF) algorithm, aiming to improve the security and stability of IoT systems by combining the advantages of deep learning and traditional machine learning algorithms. How to promote big data analysis methods to intelligent management and operational optimization is the core of the development of the IoT industry. This article ultimately proves through experiments that the AUC (area under the curve) of the CNN-RF model is closer to 1 than the CNN model, and has high classification ability.