登录
洞察 - Privacy-preserving machine learning
Differential Privacy and Its Applications in Machine Learning: Advances and Practical Implementations
Differential privacy provides a mathematically rigorous and quantifiable notion of privacy that enables high-utility data analysis while protecting individual privacy. This survey discusses recent advances in differential privacy theory, including novel variants and mechanisms, as well as the theoretical foundations and practical implementations of differentially private machine learning.
Balancing Privacy and Utility in Face Recognition Systems: A Novel Information-Theoretic Approach
This research introduces a novel information-theoretic approach to address the trade-off between privacy preservation and utility in face recognition systems. It proposes the Discriminative Privacy Funnel (DisPF) and Generative Privacy Funnel (GenPF) models to quantify and mitigate privacy risks while maintaining high-quality data analysis.
