out of memory error during key generation

Michael Roth mroth at nessie.de
Fri Jun 18 02:36:37 CEST 1999


On Thu, 17 Jun 1999, Dale Harris wrote:

> >   3DES       768
> >   CAST5       80
> >   Blowfish  4136
> >   Twofish   4256
> > 
> > This is the memory needed for the key schedule and therefore it is put
> > into the secure memory block.  In the long run I have to increase the
> > size of this block.
> 
> I imagine that there is documentation on this, but what are relative security
> of these algorithms?  Is CAST5 somehow less secure than Blowfish?  

This is the memory usage for *key schedule*. A cipher like one of the four
above is implemented like a pipeline. Their are many iterations on the
data necessary to encrypt the data. However, each iteration uses a
different so called `sub key'. This sub keys are calculated with less or
more complicated algorithms. Some of these algorithms use a very large
amount of memory, some not. But this memory usage says absolute nothing
about the strange of the cipher (at least on the named ciphers above).

It's not so easy to determine the strength of a cipher with a number to
compare cipher against and say cipher A is 2 times stronger then cipher B.

However, the used key length can give you some hints wich cipher may be
stronger and which is not so strong. But this is only a part of the whole
story. You can for example say that 3DES (tripple DES) is about two times
stronger than simple DES. But you can also prove that tripple DES is _not_
three times stronger than simple DES. Comparing different alogrithm is
more complicated because you can't look on the algorithm and say: "Hey,
this algorithm has a quality faktor of 1234". Cryptography is a very
complicated and heavy sience.

However, you asked for the key length of the ciphers, so here is a list.
Please note my above comments and don't rely to much on the algorithm key
length. (As you can note, Blowfish and CAST5 have the same key length
but the memory usage of the key schedule algorithms are very different.
The memory usage doesn't say anything about the security of an algorithm 
at all.)


        3DES         168 (really used key length; raw length is 192)
        CAST5        128
        Blowfish     128
        Twofish      256


cu
		Michael




More information about the Gnupg-devel mailing list