EvaluatingDPML
What it is
A research framework designed to evaluate the privacy leakage of machine learning models trained with differential privacy. It adapts and extends membership inference attack methodologies to assess how well various DP algorithms protect sensitive data.
The problem it solves
The difficulty of accurately measuring and auditing the privacy leakage of machine learning models, especially when using complex differential privacy mechanisms.
What you could use it for
- Evaluating the privacy guarantees of differentially private machine learning algorithms.
- Conducting membership inference attacks to test data leakage.
- Analyzing attribute inference attacks in realistic scenarios.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.