FPGrowth Parameter Tuning to avoid identifying people when attributes used in example set

Is there any guidance on how to tune FPGrowth to provide "some" degree of protection against creating association rules from data which includes peoples responses to survey questions? 

If I want to generate rules that included at least "k" people in any 1 association rule. Is it enough to set min items per itemset to "k" and does it make more sense to use frequency or support as the measure?