Sarve: synthetic data and local differential privacy for private frequency estimation

Cybersecurity

Table 6 The lowest MSE for the existing method and the enhancements which were tested on a synthetic 10-dimensional dataset having 50,000 records

Method	ε = ln(2)	ε = ln(3)	ε = ln(4)	ε = ln(5)	ε = ln(6)	ε = ln(7)
RS + FD[ADP]	0.000496	0.000255	0.000198	0.00015	0.000114	0.000111
RAPPOR	0.000807	0.00048	0.000325	0.00026	0.000247	0.000226
Hadamard response	0.000943	0.000312	0.000268	0.000215	0.000187	0.000151
Sarve	0.000481	0.000261	0.000207	0.000145	0.000118	0.000106

Method	ε = 2	ε = 3	ε = 4	ε = 5	ε = 6	ε = 7
RS + FD[ADP]	9.65E−05	4.51E−05	3.38E−05	2.45E−05	2.43E−05	2.35E−05
RAPPOR	0.000205	0.000112	6.63E−05	4.86E−05	3.06E−05	2.43E−05
Hadamard response	0.000177	0.000111	0.000126	0.000118	0.000102	9.77E−05
Sarve	0.000101	4.39E−05	3.14E−05	2.68E−05	2.41E−05	2.26E−05

The values in bold indicate privacy conditions when Sarve performed better than adaptive RS+FD and resulted in lower MSE between real and post-privatization estimated frequencies