Using machine learning methods to detect phishing emails: algorithms and approaches

The article discusses the approach to detecting phishing emails using machine learning methods. The focus is on building and comparing classification models such as logistic regression, case forest, and gradient boosting. The main focus is on how machine learning models can be built into various levels of security systems, from network protocols to applications and storage systems. Disclosed is a method of preprocessing text data, involving extraction of key features. Analyzed the main measures of protection against mail phishing. Methods for selecting features to increase the accuracy of models based on filtering methods, wrapper methods and embedding methods are analyzed. The article formulates the requirements for the development of an anti-phishing system, including taking into account the linguistic features of the text. A prototype of software has been developed, which involves the introduction of a comprehensive hybrid approach to content analysis based on machine learning algorithms. The software allows you to flexibly shape the user's behavior strategy in the event of a deliberate phishing attack.