Skip to main content

Table 7 The lowest MSE for the existing method and the enhancements, tested on synthetic 20-dimensional dataset having 50,000 records

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

Method ε = ln(2) ε = ln(3) ε = ln(4) ε = ln(5) ε = ln(6) ε = ln(7)
RS + FD[ADP] 0.000105 6.24E05 4.30E05 3.35E05 2.77E−05 2.41E05
RAPPOR 0.000155 0.000121 9.77E−05 8.32E−05 7.35E−05 6.61E−05
Hadamard response 0.000159 7.82E−05 5.37E−05 4.32E−05 3.57E−05 3.27E−05
Sarve 0.000113 6.46E−05 4.51E−05 3.39E−05 2.77E−05 2.50E−05
Method ε = 2 ε = 3 ε = 4 ε = 5 ε = 6 ε = 7
RS + FD[ADP] 2.35E−05 8.77E−06 4.36E−06 2.69E−06 2.17E−06 1.78E−06
RAPPOR 6.64E−05 3.93E−05 2.24E−05 1.31E−05 8.06E−06 5.27E−06
Hadamard response 3.29E−05 2.12E−05 1.90E−05 1.77E−05 1.58E−05 1.75E−05
Sarve 2.27E05 8.76E06 4.33E06 2.55E06 1.96E06 1.72E06
  1. 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
  2. The attribute values followed non-uniform distribution