Not exactly the same, but some folks on my old team at Google spent a long time trying to get differential privacy for ML to work in prod but the level of noise it added to everything was just incredible, I would highly recommend staying away from the technology unless you have no other choice (i.e. people will just not give you data otherwise)