Learning With Differential Privacy.
Privacy-Preserving Machine LearningSpringerBriefs on Cyber Security Systems and Networks(2020)
摘要
AbstractConsidering the internal representations in machine learning models may potentially imply some of training data, an adversary may launch attacks to extract parts of training data from a trained model (e.g., the “black-box” model-inversion attack Fredrikson et al. (2015)). Thus, the privacy guarantees if datasets contain correlated inputs is also important.
更多查看译文
关键词
differential privacy,learning
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要