I think, Andrey makes a very important point here. The option to use 3DES
symmetric encryption, SHA1 digest and ZLIB compression must remain open until
a formal process of phasing them out is initiated, with a clear road map.
Right now, excluding these algorithms would break interoperability in a very
bad way, as described by Andrey.
Of course, SHA1 and 3DES are not without problems, but for most
security-critical applications they are still perfectly adequate.
Also note that prior to ECC, any symmetric cipher could have been matched
with any public key algorithm, because the secure block size for asymmetric
encryption algorithms (ElGamal and RSA) well exceeded the key sizes for
symmetric block ciphers. The encryption of session keys with random padding
has never been an issue. With ECC, this is no longer the case. For instance,
256-bit AES keys won't fit into one ECC ElGamal blocks, which are otherwise
reasonably secure. Heck, even 3DES might be a little problematic, if 192-bit
curves are used.
Finally, I would like to draw the group's attention to a special need of
moblie communication, that seems not to receive due attention. While it is
true that computational power is less and less of an issue with mobile
phones (although there are still plenty of under-powered devices in use and
even in production), the message size issue is here to stay for much longer.
In GSM networks, which are by far the most common around the globe,
peer-to-peer messaging between handsets is done in 1120-bit chunks, for
which operators charge money. Using an RFC4880-compliant format, even the
shortest message using reasonably strong asymmetric encryption (1024-bit
RSA), requires two chunks, which cost exactly twice as much for the sender
(and in some North-American setups even the recipient) as a regular text
message. Short public key and encrypted session key sizes of ECC may
actually prove to be the primary driving force behind its adoptation in the
mobile world, even if Moore's law renders low computational costs
irrelevant.
On Fri, Feb 29, 2008 at 09:36:11PM -0800, Andrey Jivsov wrote:
David Crick wrote:
...
The ecc-pre-6.txt's section 2 pretty much says "this is how to
do ECC with AES," and we've said that this proposal is a "MAY."
In a sense this is therefore some kind of a fork (sub-superset?)
of 4880, so we're not concerned with 3DES (or CAST), and - as
with DSA - we can make specific restrictions in order to meet
compliance.
I think we need to be careful here. Let's examine a use case.
I am a user of some RFC 4880 OpenPGP application. All algorithms are
available to me, e.g. I am not a government employee.
* I use the application to encrypt mail to 5 recipients, my friends.
They use RSA/DSA/ElGamal keys.
* I upgraded the application to the next version that happened to
implement "ECC in OpenPGP".
I assume that we agree that if I encrypt to exactly the same 5 people
with new version, as far as algorithm selection is concerned, the result
is identical to the previous version.
* I added 6th recipient to the list, which uses ECDH key.
I hope we will agree that it must be possible to send single e-mail to 6
recipients. RFC 4880 specifies that the default encryption algorithm is
3DES, thus, there is always a match. Why shouldn't single e-mail be sent
to 6 recipients with 3DES symmetric key?
If the application is operating in Suite-B mode, or FIPS more, etc, then
the situation is different.
Dependant on the "low spec" argument/evidence, we could just
make one of the larger curve-hash pairs as the MUST, and make
AES256 as the must cipher. Then if we need to encrypt to one
of the two bigger curves *and* a smaller curve then we just
"up" the cipher to AES256 for the small curve recipient.
RFC 4880 currently grants to a user of embedded device a method to say
"don't do AES-256 to me" by using cipher preferences that exclude AES
256. We should respect this. We cannot change the fact that there are
hardware platforms that don't implement or don't like AES 256.
signature.asc
Description: Digital signature