Abstract
This led to the development of graphical password authentication systems which are utilized as an alternative to traditional text-based authentication systems, which prove to be more secure as well as ease of use by human cognition abilities utilized to create complex passwords in a more memorable way. Updating the dataset of systems provides an avenue for future research into performance evaluation of VVSST and validation on multiple systems to address the challenges for security vulnerabilities and usability issues. In this research, the performance effect of some machine learning models has been investigated on graphical password authentication systems. We provide a comparative analysis using supervised learning approaches like Support Vector Machines (SVM), Decision Trees, and Neural Networks. Robustness and usability of these models were evaluated by analyzing the well-known performance metrics including accuracy, precision, recall, and false-positive rates. The experimental results show considerable performance variation between the models, unveiling unique strengths and weaknesses of each approach for balancing security and user experience. It enables the design of more secure and efficient graphical password authentication systems, leading to the future directions of authentication technologies.