Comment by elif
3 years ago
Perhaps the best way to build trust in a cryptographic algorithm is to have it devised by certifiably neutral general purpose mathematic neural net.
It could even generate an algorithm so complicated it would be close to impossible for a human mind to comprehend the depth of it.
> It could even generate an algorithm so complicated it would be close to impossible for a human mind to comprehend the depth of it.
Okay... then some nefarious actor's above-human-intelligence neural network instantly decodes the algorithm deemed too complicated for human understanding?
I don't see how opaque neural nets are suddenly going to make security-through-obscurity work.
"Certifiably neutral"
So, by a process that hasn't been designed yet. Especially when one considers how opaque most neutral nets are to human scrutiny.
I mean, if the source, training data, and query interface are public, it would be insanely difficult to hide a backdoor
There i "designed" your impossible criterion in just a few obvious steps you could have inferred
There are many, many papers that show how you can make innocuous changes to inputs to make neutral nets produce the wrong result. You might be overestimating the difficulty of this process.
1 reply →