Leo

Introduction本文主要贡献 : quantify membership information leakage through the prediction outputs of machine learning models.方法 : turn machine learning against itself and train an attack model

RefKairouz, P., McMahan, H. B., Avent, B., Bellet, A., Bennis, M., Bhagoji, A. N., … & d’Oliveira, R. G. (2019). Advances and open problems in federated learning. arXiv preprint arXiv:1912.04977...

k-anonymity 、k-map 、l-diversity 、δ-presence 的介绍

加性切诺夫界、乘性切诺夫界 ：Azuma不等式 ：斯特林公式 ：斯特林公式 ：斯特林公式 ：斯特林公式 ：斯特林公式 ：斯特林公式 ：...

满足差分隐私的经验误差最小化方法

标题

标题

II. DIFFERENTIAL PRIVACY AND ITS FLAVORS讨论各个差分隐私机制的优缺点

Differential privacy for machine learning models can be obtained in four ways: input perturbation, output perturbation, objective perturbation, and change in optimization algorithm.

标题

Problem Definition and Preliminaryprivacy concernsdata privacy : when the data contributors release their own data to the databasemodel privacy : when the database publishes the learned model to t...

记录几条疑问Introduction主要工作简介：

补充知识 : meta-path     \quad\quad\quad\,\,\,\,\, Graph Embedding 之 metapath2vec场景 : Attackers create fake accounts and interact with target users. (in order to dig out the user’s friend lists or social graph)攻克点 : although differential privacy has g..