ENHANCING TWO-FACTOR AUTHENTICATION FOR CLOUD HEALTHCARE SECURITY: THE ROLE OF DEEP LEARNING

Abstract

In today's digital world, secure user authentication is crucial, particularly in cloud-based environments such as healthcare, finance, and business platforms. Two-factor authentication (2FA) is gaining popularity because traditional password-based systems are becoming increasingly vulnerable to hacking. Biometric techniques, such as facial recognition, improve security; however, current solutions frequently rely on computationally demanding models and necessitate extensive datasets, which constrain their use in lightweight or cloud settings.

Building on this context, this study tackles the problem of creating a 2FA system that is both efficient and flexible, using face recognition and one-time password (OTP) verification, and that works well with limited data and computing power. The main goal is to create a useful 2FA framework based on deep learning that is simple to train, deploy, and add to modern web infrastructures.

To achieve this efficient approach, a small set of facial images is used to train a custom convolutional neural network (CNN) using triplet loss. The model creates embeddings that use cosine similarity to tell the difference between real and fake inputs. FastAPI is used to build the authentication system, which has a web-based interface for uploading images and verifying one-time passwords. During training, you can see loss and accuracy metrics in real time.

The results show that the model can generalize well with only 5-10 training samples while still having a high verification accuracy. This architecture is ready for the cloud, is light, and can be used in real-world secure applications. Future improvements might include being able to add more users, using the camera, and connecting to cloud-based identity providers.