Factorization of RSA-180

[Robert Holmes]

This just showed up on ePrint:

"Factorization of RSA-180" - http://eprint.iacr.org/2010/270

Another large factorization possible by ggnfs+msieve.

[moebius]

I have done a Polynomial selection for RSA-180 at Jan 19th 2010 with msieve_gpu.


results for example


# norm 3.238498e-018 alpha -5.244691 e 2.982e-014
skew: 154808464.30
c0: 278552759242679896359919601738676051354776448
c1: -48023424149645535510809863817782915964
c2: -263808952653699707784808442634
c3: 291711014051391544978336
c4: 9424826240603
c5: 4320
Y0: -134640945149253025253619248206651255
Y1: 12665893392257695477

.
.
.

[Jens K Andersen]

It seems odd that the paper doesn't give the factorization. I wonder whether it is deliberate or they forgot this rather important detail in their eager to not be scooped again. As I mentioned in another thread, an anonymous Wikipedia edit does give the factorizarion and credits them: "RSA-180 was factored on May 8, 2010 by S. A. Danilov and I. A. Popovyan from Moscow State University, Russia."

[Batalov]

Weren't they essentially scooped back in Nov.2008?
That was a home computer-class excercise, too; pol51, msieve, volunteers, + coffee + time = factorization.

[R.D. Silverman]

Originally Posted by Jens K Andersen

It seems odd that the paper doesn't give the factorization. I wonder whether it is deliberate or they forgot this rather important detail in their eager to not be scooped again. As I mentioned in another thread, an anonymous Wikipedia edit does give the factorizarion and credits them: "RSA-180 was factored on May 8, 2010 by S. A. Danilov and I. A. Popovyan from Moscow State University, Russia."

I might have been impressed if they had WRITTEN AND USED THEIR OWN
CODE. As it is, it is just a mostly mindless exercize in using someone else's
code. It is not worth publishing.

[jasonp]

They are in the middle of adding MPI support to the block Lanczos code, which I have been wanting to start for months now.

It's an interesting question whether RSA-190 is accessible now in the same way that RSA-180 was.

[wblipp]

Originally Posted by R.D. Silverman

I might have been impressed if they had WRITTEN AND USED THEIR OWN CODE. As it is, it is just a mostly mindless exercize in using someone else's code. It is not worth publishing.

And yet newsworthy. Is the problem a lack of channels for newsworthy but not publishable events, or the failure to use such channels in preference for the more prestigious channels?


P.S. You fool no one. You would not have been impressed of they had written their own code implementing someone else's algorithm, either. Code writing is not the key to garnering your admiration.

[Raman]

Originally Posted by jasonp

They are in the middle of adding MPI support to the block Lanczos code, which I have been wanting to start for months now.

It's an interesting question whether RSA-190 is accessible now in the same way that RSA-180 was.

Once that this has been done entirely, we shall do M1061 together over within this forum. I will be ready at first of all, in order to run up the LA for that exactly, at once

[fivemack]

I'm confident that RSA-190 is accessible; at least, I believe the software will run 33-bit LA, I have a 32G machine that can do the matrix-creation step, I believe boinc is large enough to do the sieving, and after less than two weeks on a GTX-275 I had a reasonable polynomial (E=9.253e-15, yield (16e, alim=rlim=2e8) 2.4486 for q=3e8..3e8+1e4 at 1.06470 sec/rel) for the larger C192 from 2^1087-1 that Bos killed off by ECM in February.

I have no desire to do RSA-190; to spend a kilowatt-decade of electricity (ten thousand dollars!) on breaking apart something that we know can be broken apart and of whose constituents we have no interest is purely wasteful.

[ixfd64]

Originally Posted by R.D. Silverman

I might have been impressed if they had WRITTEN AND USED THEIR OWN
CODE. As it is, it is just a mostly mindless exercize in using someone else's
code. It is not worth publishing.

That's it, I'm quitting GIMPS right now. Since almost everybody uses Prime95 instead of writing their own program, discoveries of new Mersenne primes are obviously not newsworthy.

[Raman]

Originally Posted by fivemack

... and after less than two weeks on a GTX-275 I had a reasonable polynomial (E=9.253e-15, yield (16e, alim=rlim=2e8) 2.4486 for q=3e8..3e8+1e4 at 1.06470 sec/rel) for the larger C192 from 2^1087-1 that Bos killed off by ECM in February.

Killed? The number still exists, as yet. There is no harm in doing 2,1087- c192 by using GNFS even now (in order to check out if that same factors are returned). Only thing is that its factors are being known right now.

Originally Posted by ixfd64

That's it, I'm quitting GIMPS right now. Since almost everybody uses Prime95 instead of writing their own program, discoveries of new Mersenne primes are obviously not newsworthy.

That's the problem with implementing with our own code. They will in no way comparable in speed with the world class one. Even by a factor of 10 to 100. Plenty of optimizations to be worked upon, what is the point in another person rewriting that same algorithm, that has already been implemented up by someone else by using their effort? It will take a while for me to understand about how that Fast Fourier Transform or that Discrete Weighted Transform works out, which right now plays a major roll with the ongoing Lucas Lehmer Tests. Even then, the 100% contribution cannot be from a single person at all, since one have to be dependent upon that GMP or some other library, which is surely, entirely written up by someone else (even especially as a team) only actually, after several months of man-work, rather.