Intelligent Malicious URL Detection Using Kernel PCA-SVM-GA Model with Feature Analysis

Dangerous websites, or rogue URLs, pose a serious threat to cybersecurity. Malicious URLs cost businesses and consumers alike billions of dollars annually because the system lure unsuspecting users into scams that steal personal information, cause financial loss, or install malware. Immediate action is required to detect and eliminate these threats. This form of detection has traditionally relied on blacklists. A major drawback of blacklists is that the system are not comprehensive and do not detect newly established harmful URLs. The procedure consists of three stages, which are preprocessing, feature extraction, and training the model. Important parts of data preprocessing when dealing with URLs are standardization and parsing. When it comes to detecting dangerous URLs using machine learning, feature extraction is a major concern. The proposed approach used a Kernal PCA-SVM-GA to train the model. The proposed approach beats both SVM and GA, with an average accuracy of 93.52%.