Unifying Re-Identification, Attribute Inference, and Data Reconstruction Risks in Differential Privacy

Jamie Hayes (Google DeepMind) · Georgios Kaissis (Hasso-Plattner Institute) · Flavio Calmon (Harvard University) · Bogdan Kulynych (Lausanne University Hospital) · Juan Gomez (Harvard University) · Borja Balle (DeepMind) · Jean Raisaro (CHUV - University Hospital Lausanne)
$\varepsilon$-dp$f$-dpaccuracy increaseattack success boundsattribute inferencebaseline riskconcentrated dpdata reconstructiondifferential privacyhypothesis testingnoise calibrationr\'enyi dpre-identificationrisk evaluationtext classificationtunable privacy

Differentially private (DP) mechanisms are difficult to interpret and calibrate because existing methods for mapping standard privacy parameters to concrete privacy risks