arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1502.07645stat.MLcs.LG
收藏

隐私免费:后验抽样和随机梯度蒙特卡罗

Privacy for Free: Posterior Sampling and Stochastic Gradient Monte Carlo

Carnegie Mellon University(卡内基梅隆大学);Marianas Labs Inc.(马里亚纳斯实验室公司)

Yu-Xiang Wang, Stephen E. Fienberg, Alex Smola

摘要:

我们考虑了敏感数据集上的贝叶斯学习问题,并提出了两个简单但有点令人惊讶的结果,将贝叶斯学习与“差分隐私:”联系起来,这是一种在保护数据库级实用程序的同时保护个人级隐私的加密方法。具体而言,我们表明,在标准假设下,从后验分布中获得一个单一的样本是“免费”的差异私有。我们将看到,只要我们感兴趣的贝叶斯模型是一致的、最优的和易处理的,估计量就是统计上一致的、接近最优的和计算上易处理的。类似地,但单独地,我们表明,最近的一行作品,使用随机梯度混合蒙特卡罗(HMC)采样也保持差异隐私与微小或根本没有修改的算法程序,这些意见导致隐私约束下的贝叶斯学习的“随时”算法。我们证明了它比最先进的差分私有方法在合成和真实数据集上的性能要好得多。

英文摘要:

We consider the problem of Bayesian learning on sensitive datasets and present two simple but somewhat surprising results that connect Bayesian learning to "differential privacy:, a cryptographic approach to protect individual-level privacy while permiting database-level utility. Specifically, we show that that under standard assumptions, getting one single sample from a posterior distribution is differentially private "for free". We will see that estimator is statistically consistent, near optimal and computationally tractable whenever the Bayesian model of interest is consistent, optimal and tractable. Similarly but separately, we show that a recent line of works that use stochastic gradient for Hybrid Monte Carlo (HMC) sampling also preserve differentially privacy with minor or no modifications of the algorithmic procedure at all, these observations lead to an "anytime" algorithm for Bayesian learning under privacy constraint. We demonstrate that it performs much better than the state-of-the-art differential private methods on synthetic and real datasets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑