What I learned by watching cryptographers debate is that attack vectors can be sufficiently estimated with some good certainty, but proving their assessments is generally impossible.