On Differentially Private Inductive Logic Programming

Chen Zeng, Eric Lantz, Jeffrey F. Naughton, David Page

Abstract: We consider differentially private inductive logic programming. We begin by formulating the problem of guarantee differential privacy to inductive logic programming, and then prove the theoretical difficulty of simultaneously providing good utility and good privacy in this task. While our analysis proves that in general this is very difficult, it leaves a glimmer of hope in that when the size of the training data is large or the search tree for hypotheses is “short” and “narrow,” we might be able to get meaningful results. To prove our intuition, we implement a differentially private version of Aleph, and our experimental results show that our algorithm is able to produce accurate results for those two cases.