Partial sensitivity analysis in differential privacy
This work addresses the need for more granular privacy analysis in differential privacy, offering a method to understand feature-specific impacts, though it is incremental as it builds on existing individual Rényi DP frameworks.
The paper tackled the problem of quantifying how each input feature contributes to an individual's privacy loss in differential privacy, introducing partial sensitivity to compute feature-level influences on gradient norms and demonstrating its application on private database queries and neural network training.
Differential privacy (DP) allows the quantification of privacy loss when the data of individuals is subjected to algorithmic processing such as machine learning, as well as the provision of objective privacy guarantees. However, while techniques such as individual Rényi DP (RDP) allow for granular, per-person privacy accounting, few works have investigated the impact of each input feature on the individual's privacy loss. Here we extend the view of individual RDP by introducing a new concept we call partial sensitivity, which leverages symbolic automatic differentiation to determine the influence of each input feature on the gradient norm of a function. We experimentally evaluate our approach on queries over private databases, where we obtain a feature-level contribution of private attributes to the DP guarantee of individuals. Furthermore, we explore our findings in the context of neural network training on synthetic data by investigating the partial sensitivity of input pixels on an image classification task.