SGF is a synthetic generation tool that provides a privacy guarantee based on plausible deniability.
See also this paper.
These directly correspond to configuration options passed to the sgf tools. More detail about these inputs can be found in the SGF documentation.
gamma(number): Must be > 1.0. Privacy parameter which controls how the plausible seeds for a record are determined.omega(integer): The number of resampled attributes is N - omegancomp(string): Noise distribution for the generative model, it can be None (no privacy), Laplace, or Geometric.ndist(string): The noise distribution in the generative model: "none", "lap" (Laplacian) or "geom" (Geometric).k(integer): Minimum number of plausible seeds that a synthetic data point needs to have to be released.epsilon0(number): Differential privacy parametertinc(integer): Steps size to create for trade-off curve (between 1 and k - 1).
For more detail in the parameters please refer to the README.pdf file found in the SGF package.