Privacy Preservation of EHR Datasets Using Deep Learning Techniques
Manika Manwal, Kamlesh Chandra Purohit · 2024
The privacy and security of Electronic Health Records are major concerns in the area of medicine. The focus of this paper is on applying deep learning techniques involving Variational Autoencoders (VAE), Long Short-Term Memory (LSTM) networks, and Support Vector Machines (SVM) to improve privacy preservation of EHR datasets. We are applying homomorphic encryption and federated learning over the design of these models, ensuring that sensitive information about the patients is perfectly secured without affecting the efficacy of data analysis and the results of the predictions. In this work, we present a comparison of the performance metrics for these privacy-preserving techniques with traditional methods in terms of accuracy, precision, recall, and Fl-score. LSTM has the highest accuracy of 0.88 followed by SVM with an accuracy of 0.86 and then comes V AE with an accuracy of 0.85. Results demonstrate that a combination of Deep Learning with state-of-the-art privacy techniques is likely to strengthen the security of data without necessarily sacrificing either high accuracy or robustness while analyzing EHR data. These findings may further create real opportunities for the construction of much more secure and reliable healthcare information management systems that ensure the confidentiality of patients and compliance with the various laws enacted for the protection of data. This will, therefore, contribute to the future of advanced methodologies safeguarding sensitive medical information while using the analytical power of deep learning.