Network intrusion detection is crucial for enhancing network security. However, existing intrusion detection models face three issues. Firstly, these models often prioritize overall accuracy, neglecting the precise identification of various attack types. Secondly, due to redundancy among features in complex high-dimensional attack traffic, models struggle to extract key information from vast feature sets. Lastly, when dealing with imbalanced data, models tend to learn features from classes with a larger quantity, thereby neglecting learning from classes with fewer instances. To address these issues, this paper proposes a network intrusion detection model, NIDS-CNNRF, which integrates One-Dimensional Convolutional Neural Network(1D-CNN) and Random Forest (RF). In this model, 1D-CNN is responsible for feature extraction, while RF is used for attack traffic classification, enabling precise identification of various attack types. To tackle the data imbalance issue, the Adaptive Synthetic (ADASYN) algorithm is employed to mitigate the model’s bias towards learning from classes with a larger quantity. Additionally, Principal Component Analysis (PCA) is introduced to address redundancy among features in complex high-dimensional attack traffic, allowing the model to accurately extract key information. Experimental evaluations demonstrate that the NIDS-CNNRF model significantly enhances network security compared to traditional intrusion detection models across KDD CUP99, NSL_KDD and CIC-IDS2017 datasets.