| Abstrakt: |
The option to steal a significant quantity of important information without the owner of the message's consent is provided by keyloggers, which are tools designed to record each keystroke made on the computer. Online criminals often employ malware-infected software to attack mobile devices like smartphones and tablets. In addition, hackers are becoming smarter over time. It will be easier for them to add a keylogger to a website than a software program because users must download and install it on their devices before accessing it. All these processes, however, are not required for a website. Websites can be run on any platform. They might pick any user as their target. Thus, social networking sites, internet banking, and emails are accessible to hackers. This paper aims to develop a machine learning-based model for an in-website keylogger detection for platform-independent devices to enhance internet users' privacy and security. The study employs Random Forest, LightGBM, and CatBoost as classifiers, and uses a hybrid feature selection method, known as Hybrid Ensemble Feature Selection (HEFS), which makes the identification process robust and less runtime complex. When comparing the selected and full features on the adopted classifiers, Random Forest was found to be the best in performance; it experienced a minimal accuracy deterioration of 1.59% while achieving a massive 84.5% reduction in feature space. [ABSTRACT FROM AUTHOR] |
