School of Mathematics and Statistics, Baicheng Normal University, Baicheng, China.
Comput Intell Neurosci. 2021 Dec 30;2021:7962489. doi: 10.1155/2021/7962489. eCollection 2021.
In this paper, we give a modified gradient EM algorithm; it can protect the privacy of sensitive data by adding discrete Gaussian mechanism noise. Specifically, it makes the high-dimensional data easier to process mainly by scaling, truncating, noise multiplication, and smoothing steps on the data. Since the variance of discrete Gaussian is smaller than that of the continuous Gaussian, the difference privacy of data can be guaranteed more effectively by adding the noise of the discrete Gaussian mechanism. Finally, the standard gradient EM algorithm, clipped algorithm, and our algorithm (DG-EM) are compared with the GMM model. The experiments show that our algorithm can effectively protect high-dimensional sensitive data.
在本文中，我们给出了一种改进的梯度 EM 算法；通过添加离散高斯机制噪声，可以保护敏感数据的隐私。具体来说，它主要通过对数据进行缩放、截断、噪声乘法和平滑步骤，使高维数据更易于处理。由于离散高斯的方差小于连续高斯的方差，因此通过添加离散高斯机制的噪声，可以更有效地保证数据的差分隐私。最后，将标准梯度 EM 算法、裁剪算法和我们的算法（DG-EM）与 GMM 模型进行了比较。实验表明，我们的算法可以有效地保护高维敏感数据。
