Designing New Cryptography for Non-Standard Threat Models

This is way to smart for me to even pretend to understand. What I do take away from the article is this quote:

> The first thing that you should do is clearly document your requirements and what attacks your system must protect against. Any undefined attack vector in your model should be assumed to be a vulnerability in your design.

Which I will borrow and use to wake up the (e-voting) audience later this week :)