Deep learning methods play a crucial role in enhancing the effectiveness of intrusion detection systems. This study presents a comparative analysis of seven deep learning models, including autoencoders, restricted Boltzmann machines, deep belief networks, convolutional and recurrent neural networks, generative adversarial networks, and deep neural networks. The primary focus is on accuracy, precision, and recall metrics, evaluated using the NSL-KDD dataset. The analysis demonstrated the high effectiveness of recurrent neural networks, which achieved an accuracy of 99.79 %, precision of 99.67 %, and recall of 99.86 %.The objective of the study: of this paper is to enhance the effectiveness of intrusion detection systems through a comparative analysis of the performance of various deep learning models and an assessment of their applicability in the context of dynamic network security threats.The proposed solution involves a comparative analysis of seven deep learning models to identify the most effective ones for network security tasks. This analysis aids in selecting the optimal models for specific security requirements.The evaluation methodology involves the use of the benchmark dataset NSL-KDD, which contains various types of attacks and normal connections. The key evaluation metrics are accuracy, precision, and recall.The system implementation is based on deep learning frameworks such as TensorFlow. The results of the system’s performance and their interpretation are presented in the paper.Experiments with the NSL-KDD dataset demonstrated accuracy, precision, and recall for all the deep learning models considered.The scientific novelty is the ability to obtain formal performance evaluations of various deep learning models for intrusion detection systems, taking into account their architectural features, the processing of temporal and spatial data, as well as the characteristics of network traffic and attack types.The theoretical significance is the expansion of methods for evaluating the effectiveness of intrusion detection systems through the analysis and comparison of the performance of deep learning models in the context of processing complex and high-dimensional network data.The practical significance is the application of the comparative analysis results for selecting the most effective solutions in intrusion detection systems and optimizing them for real-world operating conditions.