Randomness quality impacts reliability of differential privacy noise
Empirical Analysis of Randomness Quality in Differential Privacy Mechanisms
Cryptography and Security
Summary
Differential Privacy protects individuals by adding random noise to data, but this relies on good sources of randomness. The authors tested how different levels of randomness quality, from very strong to deliberately weakened, affect privacy mechanisms. They found that privacy tools can detect moderate flaws in randomness but detecting flaws does not always mean privacy is actually broken. This helps understand the real risks when randomness isn’t perfect in protecting privacy.
What this means in practice
- •For privacy engineers: Check randomness quality in privacy systems to detect potential deviations affecting noise generation in differential privacy implementations.
- •For security auditors: Use empirical methods to assess whether randomness flaws found statistically correspond to actual privacy risks in deployed systems.
Authors
Cesare Gerolimetto Fabrello, Valeria Rossi, Alberto Trombetta, Massimo Caccia
Abstract
Differential Privacy (DP) relies on carefully calibrated random noise to protect individual privacy in statistical analyses. While theoretical work has analyzed DP under weakened randomness assumptions, the practical consequences of entropy degradation remain poorly understood. We present a systematic empirical investigation of how randomness quality affects differential privacy mechanisms using IBM's DiffPrivLib. We introduce progressively degraded entropy sources characterized by established test suites, starting from high-quality quantum True Random Number Generators (TRNGs) and cryptographically secure Pseudo-Random Number Generators (PRNGs) down to systematically manipulated sources with controlled entropy degradation. Through repeated experiments over one million queries on a reference database and complementary statistical tests, we directly analyze empirical Privacy Loss Random Variable distributions. Our results demonstrate that DP mechanisms reliably detect deviations when approximately 1 bit in every 8 to 16 is manipulated, with detection sensitivity varying significantly between bit-level biases and temporal correlations. We demonstrate that statistical detection of distributional anomalies does not necessarily correspond to actual privacy guarantee violations.