Gaussian Privacy Protector for Online Data Communication in a Public World
Zahir Alsulaimawi · 2020
Capturing privacy requirements for the Internet of Things (IoT) devices is essential to create sufficient public confidence to adopt and promote such devices. However, data privacy-preserving has a particular issue in the era of IoT because of a wide variety of facets and aspects that makes it difficult to formulate. The objective of the proposed research is to design a novel continuous high-dimensional data release mechanism called Gaussian Privacy Protector (GPP). In such cases, GPP can prevent adversaries from mining private information from the released data while maximizing the amount of information revealed about the utility data. We utilize variational lower bounds of mutual information approximation that is implemented as supervised learning using an adversarial training algorithm. We use two datasets: MNIST (hand-written digits) and HAPT-Recognition (Human Activities and Postural Transitions’ Recognition using smartphone data) to evaluate our algorithm. Our experimental results on these datasets show that our approach is highly effective in removing private information from the datasets while allowing non-private data to be mined effectively.