Perhaps the most interesting (or painful) part of teaching is that you need to give a 90-minute lecture on a specific topic and you realize that you have no idea what that topic is about…

For me, differential privacy (DP) is definitely one such topic. I have never researched on DP in any way and I have to admit that it took me a great deal of courage (and time!) to cook up the slides deck for this lecture. But still, I believe that this is the right thing to do and I feel bad if a Waterloo student walks out of a security and privacy course without knowing what differential privacy is.

And I have to admit that although I am just scratching the surface of DP in the lecture, I am deeply impressed by its elegance. For example,

  • Why the probabilities are multiplicatively-bounded instead additive?
  • What does the “for all subsets of the range of f” part mean in the definition?
  • How to calculate sensitivity for bounded and unbounded DP given the domain of f?
  • What is the lower bound of noise we need to add?

Being a hardcore programmer for too long, I almost forgot how beautiful mathematical theorems and proofs can be.

As I prepare for this lecture, I got a lot of inspirations from online resources, in particular,

As a result, I’d like to share my lecture slides to pay back the kindness. The slides are no in way comparable to their in-depth courses in both expertize and insights, but could come in handy if you happen to have only 90 minutes and want to have some non-trivial content on this topic.

As a disclaimer, I am definitely not an expert in this area and this is part of my learning journey as well. I am sure there will be loose ends in the slides and if you happen to spot anything that does not make sense or have any comments and suggestions, please don’t hesitate to let me know and I’ll keep the slides update.