This research study explores the evaluation of an AI-based authentication system

The rise of digital technologies has fueled the demand for secure and convenient authentication methods. Traditional password-based systems are increasingly vulnerable to cyberattacks, prompting the exploration of alternatives like biometric authentication. Among these, artificial intelligence-based voice and face authentication systems are gaining significant traction due to their accuracy, non-invasive nature, and potential for personalization. In this article, we have incorporated findings from 2 scholarly articles that provide evidence on the efficient nature of AI-powered authentication.

José Capote-Leiva, Marco Villota-Rivillas, and Julián Muñoz-Ordóñez in their article expresses the thoughts on the AI-based authentication system. Computer security is a pressing global concern across economic sectors, as companies handle sensitive client and personnel information. The secure and consensual management of this data is legally mandated, and mishandling can lead to significant financial losses and complex legal proceedings. Implementing a biometric-based authentication system, utilizing a convolutional neural network like MobileNet, has demonstrated high accuracy in user classification, achieving 96% accuracy for voice and 100% accuracy for face recognition. This approach, when combined with low-cost devices like Raspberry Pi 3, provides a robust, cost-effective, and high-performance solution for accurately recognizing biometric patterns, offering enhanced security for critical organizational assets. Based on the results of the study, it can be said that deep learning neural networks developed for biometric security provide more control over authentication in the different locations where the biometric system is implemented.

Salman Baig, Kasuni Geetadhari, Mohd Atif Noor, and Amarkant Sonkar in their article in an effort to minimize errors in traditional attendance marking systems, authors have successfully implemented an automated attendance system based on deep learning. This system utilizes face recognition technology, demonstrating a high level of robustness in accurately identifying users with an impressive accuracy rate of 98.3%. Furthermore, the results of the system's recognition process are efficiently converted into a PDF file, providing a convenient and accessible format for attendance records. This innovative approach not only addresses the limitations of manual attendance marking but also showcases the potential of deep learning in enhancing the accuracy and efficiency of attendance management systems.

In the mentioned article authors with the facial recognition component underwent testing using two distinct datasets, yielding consistent accuracy levels of approximately 90% and peaking at 95%. These results surpass initial expectations and open the door for practical real-world applications. Additionally, the study showcases two tangible applications of the facial recognition technology. The first is an online web tool designed for effortless training and testing of the entire facial recognition system using image sets. Despite being an alpha version with limited functionality, it has proven to be fully operational and valuable for demonstration purposes. The second application involves identifying individuals in videos, where the system processes the video, tracks individuals with bounding boxes, and performs accurate person identification. Ongoing upgrades are underway to enable recognition of multiple individuals simultaneously.

The successful implementation of biometric-based authentication systems, coupled with the robustness and high accuracy of deep learning models, demonstrates their potential for revolutionizing security measures and attendance management. These advancements not only address the pressing global concern of computer security but also pave the way for practical applications, showcasing the transformative power of AI and deep learning in safeguarding sensitive information and optimizing authentication processes. The findings underscore the promising trajectory of these technologies in providing cost-effective, accurate, and reliable solutions for critical organizational needs.