Аннотация
The increasing complexity of attacks and the growth in network traffic volume reduce the effectiveness of traditional solutions in the field of network security. Signature-based and rule-based detection and prevention systems are unable to provide an adequate level of protection in network environments characterized by the emergence of previously unknown threats. Therefore, intelligent methods designed for the analysis of network traffic in streaming mode have become highly important from both scientific and practical perspectives. The main purpose of this review article is to conduct a comprehensive analysis of intelligent methods for detecting and preventing network attacks in a streaming data processing environment. In addition, the article systematically examines machine learning (ML), deep learning (DL) and hybrid approaches based on continuous real-time analysis of network traffic. The article reviews recent scientific research conducted by domestic and international scientists, also analyzing the advantages, limitations and practical applicability of the employed intelligent methods. The architecture of intrusion detection and prevention systems (IDS/IPS) is also examined, the highlighting of integrating intelligent approaches for the advancement of network security. The results of the analysis indicate that machine learning methods are effective in resource-constrained environments, whereas deep learning techniques achieve higher accuracy in detecting complex and previously unseen attacks. Hybrid intelligent systems emerge as a promising direction, providing an optimal balance between detection accuracy and computational efficiency.