Modeling, Quantifying, and Limiting Adversary Knowledge