Clear-Text Patch? was: Re: OpenSSH 5.1: call for testing

rapier rapier at psc.edu
Thu Jul 24 02:10:58 EST 2008



Dag-Erling Smørgrav wrote:

> Depends on the context.  I rarely use the "none" cipher, and haven't in
> a while, but it has always been on a trusted network, between two
> servers connected to the same switch.  I would never use the "none"
> cipher over an untrusted link, even if only for "bulk data transport".

There are times in which en clear authentication is acceptable but I 
find it probably better to treat every network as not being trust 
worthy. However, I do think there are many times in which is perfectly 
acceptable to transfer bulk data en clear - most data is not sensitive 
to the point where leaking it would prove to be problematic. For 
example, if you are transferring a hundred terabytes of weather station 
data to a supercomputing site for storm analysis (which is the class of 
user I write for) the data isn't sensitive at all. Primary concerns 
would be the speed of the transfer and meeting the required security 
aspects (authentication, data integrity). This is why the majority of 
the high performance bulk data transfer applications - like kftp, 
gridftp, bbftp, and so forth don't encrypt the data but use strong 
security on the authentication.

However, there are time when this is unacceptable - which is why we 
developed a multi-threaded aes-ctr cipher. We've been able to get 
900Mb/s transatlantic with aes256 that way.

Also, I'm thinking that you actually do unencrypted bulk data transfer 
over untrusted links. I know I just did it a few minutes ago when I got 
the 5.1 OpenSSH distribution from a mirror site.


More information about the openssh-unix-dev mailing list