In the context of the rapid growth of cyber threats and an increase in the volume of network traffic, the detection and prevention of network intrusions are becoming urgent tasks to ensure cybersecurity. The purpose of this study is to develop an effective Mini-VGGNet model for detecting anomalies in unbalanced network traffic. The study also aims to analyze time dependencies in the data, which allows for more accurate identification of potential threats. The work uses a deep learning methodology based on the Mini-VGGNet architecture, which includes convolution layers and pooling layers to extract features from network data. To improve detection efficiency, data preprocessing methods are used, including removing unnecessary values and normalization, which improves the quality of model training. The model is trained on a dataset containing various types of network attacks, which allows you to identify anomalies in real time and adapt to changes in network traffic. As a result of the conducted research, high accuracy in detecting network intrusions has been achieved, which confirms the effectiveness of the proposed model. The results show that the Mini-VGGNet model is significantly superior to traditional methods in terms of speed and accuracy of anomaly detection. The significance of the work lies in its contribution to the development of cyber defense methods and improving the security of information systems, which is critically important in the context of constantly changing cyber threats. The results can be used for further research in the field of cybersecurity and the development of more advanced models for detecting network attacks.