Algorithmic study on efficiently computing similarities to private datasets for improved privacy-utility trade-offs.