Privacy-Preserving AI Computation for Digital Dentistry Using CKKS-Based Homomorphic Encryption in HTTPS at Server Side
CKKS, Oral Surgery, Artificial Intelligence in Dentistry, Privacy-Preserving Computation, Cloud Security, Encrypted Web Processing
AuthorsAbstractAI-driven dental imaging and oral diagnostic systems continuously transmit encrypted patient radiographic and clinical data to cloud-based diagnostic platforms for secure intelligent analysis. While server communication is over HTTPS and protects patient dental records when it is sent to the server, once it is received by the server, patient dental records must be decrypted before any other processing occurs. At this stage, the data becomes visible in memory, which leaves them vulnerable to insider misuse, memory disclosure, or compromise of the underlying cloud infrastructure. This work explores how fully homomorphic encryption can be used to extend confidentiality beyond the transport layer. This study examines how privacy can be preserved during server-side web computation by combining standard HTTPS with the CKKS homomorphic encryption scheme. In the proposed framework, AI-assisted oral diagnostic inference and dental imaging analytics are performed directly on encrypted clinical dental datasets without exposing patient-sensitive information. Therefore, servers will have no access to their decrypted representations (or plain-text forms) throughout the entire process chain. Consequently, the privacy/security aspects of those sensitive pieces of information are still preserved, even when computations occur on any infrastructure that may not have complete trustworthiness attached thereto. The paper describes the overall client–server architecture, the associated cryptographic workflow, and the CKKS formulation required to support approximate arithmetic under realistic noise constraints. It also gives correctness conditions and noise growth, and SIMD-enabled encrypted inference allows efficient processing of panoramic radiographs, CBCT scans, intraoral images, and AI-driven oral pathology prediction workloads under strict privacy-preserving conditions. The results suggest that CKKS-based homomorphic computation can be practically combined with HTTPS to provide stronger, end-to-end privacy guarantees for modern web services. 1. INTRODUCTION Artificial Intelligence (AI) is rapidly transforming oral surgery and dentistry through intelligent diagnostic systems, automated radiographic interpretation, AI-assisted prosthodontic planning, and cloud-based dental analytics. Modern digital dentistry platforms increasingly rely on deep learning models for oral lesion detection, dental image segmentation, implant planning, restorative treatment prediction, and AI-assisted clinical decision support. These intelligent systems process highly sensitive patient information including panoramic radiographs, cone beam computed tomography (CBCT) scans, intraoral images, prosthodontic records, and restorative treatment data. To provide these services, servers require user authentication and therefore •••••••••••••••••••••••••••• ejprd.org - Published by Riset Publishing Services LLC
EJPRD
Copyright ©2026 by Riset Publishing Services LLC