Differential Privacy Analysis using Convolutional Neural Networks in COVID-19 X-ray Images Classification Model

Abstract

This paper proposed that the artificial intelligence (AD methods be used in conjunction with medical to detect COVID-19 affected individuals. But the investigations needs to be performance with maximum caution so as not to jeopardise the privacy of the person’s data that are required for these beneficial applications to be successful. As a result, the differential privacy (DP) application, is deemed to be an intriguing research topic, has been utilized in this study. For the development of a diagnostic concept, XRay pictures obtained from COVID-19-infected public data sources were utilised. This model has been trained using Convolutional neural networks (CNN), an effective deep learning model, and it proposes the probability of infection with an accuracy of 94.7 percent using CNN model. Among the most significant findings of the research was the recommendation of differentiated privacy practise PATE for such apps in order for them to be trustworthy in real-world use scenarios. According to this point of view, tests were redone with DP-applied pictures, and the findings obtained were then discussed. An method called Private Aggregation of Teacher Ensemble (PATE) was utilized to assure that the students’ personal information was protected.

Ähnliche Arbeiten