Differential Privacy using Fuzzy Convolution Neural Network (DP-FCNN) with Laplace Mechanism and Authenticated Access in Edge Computing

During recent years, mobile edge computing is getting much attention from both academia and industry. However, many found that this emerging architecture needs a proper data privacy protection mechanism at mobile edge nodes against unintended data use by authorized data analysts. Due to the reason, the development of a proper lightweight privacy-preserving data analysis mechanism is of great urgency. Thus, we propose DP-FCNN, a light-weight Differential Privacy (DP) framework using Fuzzy Convolution Neural Network (FCNN) with Laplace Mechanism which injects noise into the personal data before uploading data from users generating the data into the storage so that the data is still useful but data privacy can be properly protected against unauthorized data analysis attempt. We implemented the proposed framework, and tested its performance in terms of scalability, processing time, and accuracy. The result shows that the proposed framework is very practical.