МАШИНАЛЫҚ ОҚЫТУ АЛГОРИТМДЕРІН САЛЫСТЫРА ОТЫРЫП ЖЕЛІЛІК ШАБУЫЛДАРДЫ АНЫҚТАУ

Abstract

The development of modern information systems has increased the complexity of network attacks and increased the relevance of security issues. This article compares the effectiveness of machine learning (ML) algorithms in detecting network attacks. Various ML methods, including neural networks, decision trees, deep learning, and adversarial machine learning, are accepted as the object of research on the accuracy and speed of detecting network attacks.  The results of the study highlight the importance of adaptive and innovative machine learning solutions for network security. In addition, directions and recommendations for future research will be given, which will increase the potential and effectiveness of machine learning algorithms in combating network attacks.  This article analyzes the machine learning algorithms used to detect network attacks. Traditional methods are becoming insufficient to counter these threats, so machine learning technologies are considered an effective alternative. The article examines the methods of supervised learning, unsupervised learning and partially supervised learning. Among the methods of supervised learning, logistic regression, decision trees, random forests, auxiliary vector machines and neural networks are considered. K-Means clustering, autoencoders, and key component analysis methods as unsupervised learning methods are discussed.  Methods of feature propagation and self-learning are being studied as partially controlled learning methods. The advantages and disadvantages of each algorithm are compared and their effectiveness in detecting network attacks is analyzed.