Abstract:Privacy-preserving and secure data sharing are critical for medical image analysis while maintaining accuracy and minimizing computational overhead are also crucial. Applying existing deep neural networks (DNNs) to encrypted medical data is not always easy and often compromises performance and security. To address these limitations, this research introduces a secure framework consisting of a learnable encryption method based on the block-pixel operation to encrypt the data and subsequently integrate it with the Vision Transformer (ViT). The proposed framework ensures data privacy and security by creating unique scrambling patterns per key, providing robust performance against leading bit attacks and minimum difference attacks.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve how to maintain efficient and accurate classification performance while protecting data privacy and security in medical image analysis. Specifically, this research focuses on the following problems: 1. **Privacy Protection and Data Sharing**: - Medical data (such as MRI images and histopathological images) are at risk of privacy leakage during sharing and analysis. When existing deep neural networks (DNNs) are applied to encrypted medical data, a compromise between performance and security often has to be made. - The research proposes a framework based on Vision Transformer (ViT), combined with a learnable encryption method, to ensure the privacy and security of data during transmission and processing. 2. **Efficient Classification on Encrypted Data**: - Performing medical image classification in an encrypted state is a challenge because encryption will destroy the original structure of the image, making it difficult for existing models to extract effective features. - This research can retain the features required for classification while encrypting the image by introducing a learnable encryption method based on block - pixel operations, thus achieving efficient and accurate classification. 3. **Robustness against Adversarial Attacks**: - Encrypted medical data needs to be able to resist various attacks (such as leading - bit attacks and minimum - difference attacks) to prevent sensitive information from being maliciously reconstructed. - The framework proposed in this research ensures robustness against these attacks through a unique encryption method, further enhancing data security. ### Summary By introducing a ViT - based privacy - protection framework, this paper solves the contradiction between data privacy, security, and classification performance in medical image analysis. This framework can not only effectively encrypt medical data but also maintain high - precision classification performance and has strong anti - attack capabilities. This provides a new solution for secure medical data analysis on cloud platforms.

ViT Enhanced Privacy-Preserving Secure Medical Data Sharing and Classification

A Privacy Protection Framework for Medical Image Security without Key Dependency Based on Visual Cryptography and Trusted Computing

S-E Pipeline: A Vision Transformer (ViT) based Resilient Classification Pipeline for Medical Imaging Against Adversarial Attacks

Evaluating and enhancing the robustness of vision transformers against adversarial attacks in medical imaging

A New Data Model for the Privacy Protection of Medical Images

Privacy-Preserving Image Classification Using Vision Transformer

Heart disease data based privacy preservation using enhanced ElGamal and ResNet classifier

PRIMIS: Privacy-preserving medical image sharing via deep sparsifying transform learning with obfuscation

Implementing vision transformer for classifying 2D biomedical images

Privacy-Preserving Medical Image Classification through Deep Learning and Matrix Decomposition

Privacy Preserving Inference for Deep Neural Networks: Optimizing Homomorphic Encryption for Efficient and Secure Classification

A Deep Cryptographic Framework for Securing the Healthcare Network from Penetration

Privacy-Preserving Deep Learning With Learnable Image Encryption on Medical Images

Advancing Healthcare: Innovative ML Approaches for Improved Medical Imaging in Data-Constrained Environments

Deep neural network-based secure healthcare framework

Privacy-Preserving Deep Neural Networks with Pixel-Based Image Encryption Considering Data Augmentation in the Encrypted Domain

MedViT: A robust vision transformer for generalized medical image classification

Tamper Detection and Self-recovery in a Visual Secret Sharing based Security Mechanism for Medical Records

Investigating the Robustness of Vision Transformers against Label Noise in Medical Image Classification

A separable privacy-preserving technique based on reversible medical data hiding in plaintext encrypted images using neural network

Privacy Protection of Medical Data Based on Multi-Scroll Memristive Hopfield Neural Network