3 ms·
Crypto-newb question: is there any reason it takes the "number of computations" difficulty as a given, rather than estimating the rate at which attacks will get
by SilasX 14y ago
Crypto-newb question: is there any reason it takes the "number of computations" difficulty as a given, rather than estimating the rate at which attacks will get smarter and require fewer computations?
The analysis that's given includes estimates of decreasing computation cost from Moore's law, but no corresponding estimate of decreasing cost from patterns found in the SHA-1 hash function.
Also, why does the cost estimate focus on commodity CPUs, rather than GPUs, which are much better at the (parallel) computations that hash attacking involves?
- jpablo 14y agoHe's only providing a lower bound for this numbers, he specifically spells your concerns. He mentions that any improvement on: * number of cores per CPU * number of CPUs per server * any improvements in cryptanalysis * instruction set improvements (arm sha1, gpus) Will only make his number even scarier.
- jwegan 14y agoIt is impossible to estimate decreases from "patterns found in the SHA-1 hash function". There is a good likelihood no patterns will be found in 10 years. It is also possible (although probably unlikely) an attack could be found that utterly breaks SHA-1.
- zeeboo 14y agoI can't answer why GPUs weren't considered, though I'd guess because they weren't as familiar with the numbers and this was a "back of the envelope" calculation. The reason why I wouldn't take into account better attacks is because that's much harder to predict than Moore's law. Also, this is supposed to be more of a worst case scenario, so even if no advances are found it will cost at least that much.