Some additional comments on this subject:

1. NIST has published its own table of equivalent strengths of various
sizes of keys and other parameters.  It's got essentially the same columns
as Lenstra's table, but doesn't have a row for every year.  Rather, it
has rows for different discrete key sizes.

It was published in IEEE Security & Privacy (a periodical) in the
March/April 2003 edition, page 48.  A PDF file is available to IEEE
subscribers, but AFAIK is not available for free on the web.

That table disagrees with Lenstra's table in a number of ways, including:
- the pairing of Discrete Log field sizes and key (private value) sizes.
- the pairing of symmetric key sizes to DL field sizes.

Here is a small excerpt from NIST's table:

  symmetric alg/key  hash    field size   DL private value size
  -----------------  ------- ----------   ---------------------
  SkipJack - 80      SHA-1     1024        160
  3DES     - 112               2048        244
  AES      - 128     SHA-256   3072        256
  AES      - 256     SHA-512  15360        512

I point this out only to show that no single table is "gospel" truth.

2. Andrey pointed out that these tables all list, in each row, a single
field-sizez value that applies to RSA, DSA and DH.  On that basis, he
suggests that NSS should employ the same limit for each of those algorithms.

However, the point of the limit imposed by NSS is to bound computation time,
and the time taken by these different algorithms for the same key/field
sizes are quite different.  Therefore, the size limits are not likely to
be equal in number of bits allowed, but rather in computational time
permitted.

That's why I wrote that the new limit on DH field sizes would be NO MORE
than (implying, possibly (much) less than) 2236 bits.

/Nelson
_______________________________________________
mozilla-crypto mailing list
[EMAIL PROTECTED]
http://mail.mozilla.org/listinfo/mozilla-crypto

Reply via email to