...

The maximum number of the unique levels of a categorical variable that are allowed to be returned to the client. If nfilter.levels is set to 0.33 (its default value), and if a categorical variable (i.e. factor) has X distinct categories then if X is greater than the 33% of the variable's length then the categories (i.e. levels) are not returned to the client. This disclosure filter protects against the disclosure of all the unique values in a numerical variable when it is converted to a factor variable. This option has been deprecated.

nfilter.levels.density

???? The maximum proportion of unique levels of a categorical variable with respect to the number of that variables that is regarded as non-disclosive. Default value is 0.33.

nfilter.levels.max

????The maximum number of unique levels of a categorical variable that is regarded as non-disclosive. Default value is 40.

nfilter.noise

The minimum level of noise that can be added to a server-side vector. The "noisy" vector can then be returned to the client. This value specifies the variance of the added noise. If nfilter.noise is set to 0.25 (its default value) then noise following a distribution (usually Gaussian) with zero mean and variance equal to the 25% of the true variance of the vector of interest is added to each individual value of that vector.

...