Re:Google AI invents its own cryptographic algorithm
by contravariant on Friday October 28, 2016 @11:57AM
(#12817374)
Cryptographical transformations aren't supposed to be one-way. Not if you know the key anyway. It would be pretty pointless to encrypt something in a way that can't ever be read.
Re:Google AI invents its own cryptographic algorithm
by andorov on Friday October 28, 2016 @03:17PM
(#12819689)
It is useful in cases where you want to verify something is the same as a stored (encrypted) value - e.g. passwords.
One-way encryption where the entire process including the equivalent of salt/pepper is publicly available and not reversible might have some interesting applications.
Re:Google AI invents its own cryptographic algorithm
by pizza on Friday October 28, 2016 @11:56PM
(#12822568)
Not that this is relevant to OP, but there are also non-differentiable descent (edit: NOT descent, but rather just general optimization) methods; https://optimization.mccormick.northwestern.edu/index.php/No...
Most importantly, f is differentiable. It has to be, since it was trained with gradient descent.
So if you want to decipher a ciphertext Y, then use backpropagation to find the X that minimizes distance(f(X), Y). You already know dF/dX (f is differentiable after all), so run this through your magic optimization solver and you'll get an X very close to the result.
Researchers use this "backproagation in input space" idea to recover the training set of a trained neural network, or to debug their model by finding inputs that activate certain neurons, for example. (This is the basic idea behind Deepdream from last year)
It's not even correct to call this "cryptography." The transformation isn't one-way.