Optimal Noise-Adding Mechanism in Differential Privacy

The authors derive a class of optimal noise probability distributions for noise-adding mechanisms for single real-valued query function to preserve differential privacy under a utility-maximization/cost-minimization framework. The class of optimal noise probability distributions has staircase-shaped probability density functions which are symmetric, monotonically decreasing and periodically decaying. In particular, they derive the optimal noise probability distributions with minimum expectation of noise amplitude and power, respectively, and compare the performances with the state of art Laplacian mechanism.

Provided by: University of Idaho Topic: Security Date Added: Dec 2011 Format: PDF

Find By Topic