Another good post. I thought that reduction-based security proofs were supposed to save us from this kind of thing. What happened? I also got the impression, back when I was into this stuff, that these proofs are quite hard to formalize. I never understood why. Otherwise, maybe model checking could find such attacks.