The main task of any company security system is to protect data, whether it is the data of the organization itself, its partners or users of its products. There are many ways to counter cyber threats, but a major drawback is the key role of human eror in these processes. Even the most diligent security team cannot get away with simply updating systems and fixing bugs in them, because threat actors are constantly finding new ways in. And when an attacker does infiltrate a system, the malware's execution speed often outpaces that of the security teams tasked with responding to them. This is not a problem that can be solved simply by hiring more people, because when the threat moves at machine speed, the response effort must be the same. Faced with this challenge, organizations are increasingly relying on machine learning products to perform cybersecurity functions that can operate autonomously and with high efficiency. The purpose of the paper is to investigate the potential application of machine learning in cybersecurity. The work described the subject area and the basic concepts, tactics, and techniques used by attackers in cyberattacks. The type of attack that poses the greatest threat at the moment, whose detection methods are flawed, a review of the machine learning algorithms used to predict this type of attack, a review of studies examining how these algorithms work with different data sets and in different combinations with each other were highlighted. Metrics for assessing the quality of the model's performance were described, and work was performed to examine and prepare the data for training and testing the model. Based on the research, a machine learning model using a gradient binning algorithm was implemented, which showed high performance on metrics such as precision, recall, accuracy, F1-score.