Abstract
Internet technologies develop very fast today. It resulted in essential growth in different types of cyberattacks and cyberthreats. Cybercriminals are becoming more advanced and utilize various techniques to access users’ confidential and financial data, avoiding their defense systems. Among these types of attacks, phishing attacks are among the most dangerous as they lead users to malicious websites where criminals can easily steal their data and credentials. Although traditional defense methods allow protection from some of these threats, they are becoming less effective due to the development of the Artificial Intelligence field. Three types of neural networks, including convolutional, recurrent, and a combination of convolutional and recurrent neural networks, were applied for protection against phishing. In the experimental results, the dataset of 392924 legitimate and 146422 malicious URLs has been used. These neural networks have received accuracy scores of 0.613, 0.719, and 0.841. The combined model showed the best scores among all models.