Malicious URLs are a serious challenge in cybersecurity, given the increasing number of threats such as malware, ransomware, spyware, phishing, defacement and trojans. Deep learning has the ability to learn complex patterns in data automatically and effectively, so it can be used to detect anomalies and malicious patterns in URLs. Previous research has proposed various methods to detect malicious URLs, including blacklist-based methods and URL features. However, these methods often lack effectiveness in dealing with evolving attack patterns. In the detection of harmful URLs, according to various studies, applying deep learning has the potential to increase the process’s efficiency and accuracy, but there is still an opportunity to further optimize efficiency and accuracy. This paper aims to develop a malicious URL detection system using deep learning based on feature extraction. This method will improve data representation through text analysis and transformation of such data, as well as selection of importan