Secure medical chatbot: leveraging encryption for patient data privacy

• Main Author: Oh, Terence Yuan Zhang
• Other Authors: Lim Wei Yang Bryan
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Other
• Online Access: https://hdl.handle.net/10356/181075
• Institution: Nanyang Technological University

In the digital era, providing accessible healthcare advice while ensuring patient data privacy has become a critical concern. This project introduces MediBot, a secure medical chatbot interface designed to offer patients reliable medical guidance using Large Language Models (LLMs), without compromising the confidentiality of their sensitive information. Recognizing the limitations and privacy risks associated with generic chatbots like ChatGPT—particularly regarding data handling and compliance with healthcare regulations—MediBot is developed as a dedicated platform tailored for hospital use. To address the challenges of securing data transmission and storage, MediBot implements a hybrid encryption approach, combining RSA and AES algorithms. This method ensures robust encryption of patient interactions, with RSA securing the key exchange and AES providing efficient data encryption for real-time communication. The chatbot is built using Flask, offering full control over both front-end and back-end components, which facilitates the integration of custom security measures and a user-friendly interface. The project explores the feasibility of training models on encrypted data and concludes that, while Homomorphic Encryption and Differential Privacy offer strong theoretical privacy guarantees, they are currently impractical for real-time applications like MediBot due to significant computational overhead and performance constraints. Instead, the focus is placed on securing data during transmission and at rest.