Differentially Private E-Values
This work addresses privacy concerns in statistical applications using sensitive data, such as healthcare, by enabling safe release of e-values, though it is incremental as it builds on existing e-value and differential privacy concepts.
The authors tackled the problem of sensitive data leakage through e-values in statistical inference by proposing a framework to transform non-private e-values into differentially private ones, achieving strong statistical power and asymptotic equivalence to non-private versions.
E-values have gained prominence as flexible tools for statistical inference and risk control, enabling anytime- and post-hoc-valid procedures under minimal assumptions. However, many real-world applications fundamentally rely on sensitive data, which can be leaked through e-values. To ensure their safe release, we propose a general framework to transform non-private e-values into differentially private ones. Towards this end, we develop a novel biased multiplicative noise mechanism that ensures our e-values remain statistically valid. We show that our differentially private e-values attain strong statistical power, and are asymptotically as powerful as their non-private counterparts. Experiments across online risk monitoring, private healthcare, and conformal e-prediction demonstrate our approach's effectiveness and illustrate its broad applicability.