Sign into an existing account

tl,dr; No.

I wrote a system that can give relatively tight upper and lower bounds on private keys in RSA but thats as far as I've got, along with some theoretical work in ECC.

I also demoed an ultra bare-bones linear regression model that used a subset of known variables from this system to give even better estimates, sort of proof-of-concept. That worked. Gave estimates for hidden variables to within ~98%

Now I'm building a full ML model to train on the data to do it properly because some of the algorithms and equations are convergent, so having good estimates lets me obtain better estimates that essentially convert them to functions and then memoize the equations in question.

I have one that is constant time but thats a huge stretch to get working for a variety of reasons, one thats O(n log n) time, and another thats subpolynomial. All of them require either knowing a hidden variable, having a good approximation of one, or knowing certain initialization values for free parameters.

This is what the ML model is supposed to help me sort out: What combination of known variables are most predictive of the hidden variables, which is critical because the base model has on the order of 1500 known or public variables in the naive case, derived only from the public key and the root of the key. With certain transformations of those variables, the count goes way higher, into the low tens of thousands, each with their own properties, algebriac relations, and ability to predict the upper or lower bounds of variables from the unknown set.

That, combine with a few dozen congruent equations (still waiting on sufficiently precise aproximators), all scattered through four+ years of research entries and logs, means its a lot to sort out, parse, and make sense of.

I'm confident that either the new model will do exactly what I intend it to do based on the proof of concept's performance, or I'll tear it out and go to a standard regressor model, though I'm avoiding that because numpy and everything built on it is a raging dumpster fire of convoluted garbage and I rather roll my own.

The plan is not to do anything myself. The plan hinges on the pure lawfulness of releasing something to the public that 1. anyone can use for any purpose, 2. anyone will use.

And that anyone could be anyone, and everyone.

Human nature will do the rest.

We're going back to 0 A.D. baby.

the Chinese just announced a second time they have cracked 2048 RSA with a q computer yesterday.

Take that for what you will...

Of course the government desperately wants RSA hacked because then they get everyone to adopt the elliptic curve parameters they get to advise on. I guess they already get that to an extent.

PROTOTYPE DINDU NUTHING!

Curious as well on the proof of concept. What kind of mfa?

OP has good ideas. Go for it @prototype. I wish I was smart enough to code like that. As soon as someone crashes this mother fucker life can go back to normal, we can wipe the kikes out once and for all, deport/deal with undesirables, race traitors, etc. and get rid of non whites.

Fucking A,' if true.