Abstract
In recent years, powered by state-of-the-art achievements in a broad range of areas, machine learning has received considerable attention from the healthcare sector. Despite their ability to provide solutions within personalized medicine, strict regulations on the confidentiality of patient health information have in many cases hindered the adoption of deep learning-based solutions in clinical workflows. To allow for the processing of sensitive health information without disclosing the underlying data, we propose a solution based on fully homomorphic encryption (FHE). The considered encryption scheme, MORE (Matrix Operation for Randomization or Encryption), enables the computations within a neural network model to be directly performed on floating point data with a relatively small computational overhead. We consider the well-known MNIST digit recognition problem to evaluate the feasibility of the proposed method and show that performance does not decrease when deep learning is applied on MORE homomorphic data. To further evaluate the suitability of the method for healthcare applications, we first train a model on encrypted data to estimate the outputs of a whole-body circulation (WBC) hemodynamic model and then provide a solution for classifying encrypted X-ray coronary angiography medical images. The findings highlight the potential of the proposed privacy-preserving deep learning methods to outperform existing approaches by providing, within a reasonable amount of time, results equivalent to those achieved by unencrypted models. Lastly, we discuss the security implications of the encryption scheme and show that while the considered cryptosystem promotes efficiency and utility at a lower security level, it is still applicable in certain practical use cases.
Highlights
Over the recent years, machine learning algorithms, with emphasis on deep neural networks, have delivered remarkable solutions for personalized medicine, enabling customized diagnosis, treatment, and prevention [1]
We have showcased the applicability of incorporating the matrix-based homomorphic encryption scheme (MORE) encryption scheme into deep learning models by tackling three different problems: digit recognition, wholebody hemodynamic analysis, and coronary angiography view classification. e first application focused on a standard benchmarking application from the computer vision realm (MNIST digit recognition) to evaluate the feasibility of a network to operate directly on encrypted data, whereas the latter two models target clinically realistic problems
E reported results indicate that the proposed solution has great potential: (i) computational results are indistinguishable from those obtained with the unencrypted variants of the deep learning-based applications and (ii) runtimes increase only marginally. e encryption scheme incurs a reasonably small computational overhead and, importantly, allows for operations to be performed directly on floatingpoint numbers, which represents a critical property for artificial neural networks
Summary
Machine learning algorithms, with emphasis on deep neural networks, have delivered remarkable solutions for personalized medicine, enabling customized diagnosis, treatment, and prevention [1]. Since deep neural networks are entirely data-driven systems that can learn explicitly from past experiences, they are commonly used as a way to integrate the knowledge and experience of medical experts into solutions for computeraided detection (CADe). To deliver results that are sufficiently reliable to be considered in clinical routines, machine learning-based solutions have to heavily rely on available medical data records [2]. Given the fact that machine learningbased solutions require access to such sensitive information, concerns have recently been raised regarding data privacy and security [3]. The confidentiality protection laws currently adopted for the manipulation of personal data (e.g., EU GDPR and US HIPAA) demand the use of more effective privacy protection strategies
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call