Differential Privacy Compute Sample Median

I’m looking for a good resource on computing the sample mean in a differentially private way.

I can see how to calculate a count or proportion, by injecting noise from a Laplace distribution; and I could use the Laplace mechanism if I knew the global sensitivity for the sample median, but I’m not sure what that would be.

Can anyone help?

Hi @mwtichenor,

I don’t have any information on this topic, however this pdf discusses the points you have mentioned in your post:

Hope this helps :slightly_smiling_face: