ANALYSIS OF MODERN METHODS OF DETECTION OF PHISHING E-MAILS

Abstract

Phishing attacks are one of the common threats to modern cyber security. The most common method fraudsters use to send fake messages to collect data is phishing emails. However, with the development of technology and artificial intelligence, the number and complexity of phishing attacks are increasing, making detecting them difficult. The article discusses traditional and modern methods of combating phishing, particularly blocklists and signature methods, and the latest machine and deep learning approaches. The analysis of the latest research made it possible to develop a generalised algorithm (fig. 2) for the implementation of the phishing email detection system, which consists of the following steps: data collection, data pre-processing, feature selection, modelling, email classification, model updating, blocking and notification/ Machine learning makes it possible to analyse large volumes of data and detect hidden patterns, which makes these methods effective for automatically blocking phishing emails. Convolutional and recurrent neural networks are also used to analyse the text of phishing messages at the level of words and phrases. Special attention is paid to developing natural language processing methods that help better understand the context of letters and detect anomalies. Deep models allow for extracting valuable features without pre-processing the data, making them practical for detecting new attacks. The implementation of machine and deep learning methods significantly increases the effectiveness of detecting phishing emails. However, further research is needed to improve and realise the models' full potential. It is necessary to create models that can independently adapt to new threats without manual intervention, analysing new patterns and strategies of attackers. This will ensure a more effective fight against phishing threats in the rapidly changing digital environment.