On 12/22/05, John Young <jya@cryptome.net> wrote:
... Is NSA regularly cracking all encryption or not, whether or with a foolproof algorithmic tool (again and again discounted by the best and brightest of the cryptographers) is that what the Times will not write about. Or via exploitation of faulty implementation.
faulty implementations meaning side channels leaking key material. i know that Glenn at Centaur/VIA is concerned about perceived pressures to keep crypto out of processor cores. consider this rumor but i'd love to see someone follow up on this story. with cache/memory timing, differential power analysis, even acoustic side channels weakening software cipher implementations (and hardware to a lesser degree) i can't help but wonder why Intel and AMD have not deployed entropy, digests, block ciphers and Montgomery multipliers in their cores - it takes very little die space and provides a huge return. makes my inner paranoid twitch... </VIA whoring> with that said, i think it's clear that a properly designed crypto system could be considered secure. the government still uses AES256 for their top secret datum, and the NSA license of ECC could be interpreted as a vote of confidence in that PK system. (or is this just another ruse? :)
There are regular calls to trust no infosec system that is not continually monitored by experts. A lesson taught by the government pros as well. ... Impregnable mathematics gradually withered as a protective rationale with the rise of the threat of faulty implementaion, black bag jobs, keystroke loggers, insider betrayal, and the gamut of traditional security failures...
with ciphers and protocols maturing is the next frontier for cypherpunks decentralized reputation and trust metrics applied to process, persons, and systems? the world around these theoretically secure ciphers is full of holes as you describe.