I'd say that science and cryptanalysis both share the ideal of (or can be viewed as) trying to approximate Bayesian updating on a Solomonoff prior. In cryptanalysis, you have the disadvantage that your opponent is intelligent and malicious (from your perspective), but you have the advantage that the encryption process doesn't have much computing power (since ciphers have to be fast to minimize computing costs). In science, nature is not malicious, but it's not limited in computing power either. Yes, they are basically similar, but each is specialized to its domain of inquiry, so I doubt science can learn much from cryptanalysis.