Using the example of vulnerability data, this talk is about what happens when data science and security collide.
When you let a data scientist loose on security data there's a ton of things you (and they) need to think about. What you think data science is, and what you expect to get from it. Why 'insight' is hard to get. How to win the battle of caveats vs usability. And how to communicate analysis when it's used to solve operational problems or report up to management.
Tl;dr - this talk is about how easily it can all go wrong - and how to stop that happening.