Re: [radext] Proposed charter text based on IETF-115 BoF

[Alan DeKok <[e-mail]>]

On Nov 24, 2022, at 3:41 AM, Stephan Mueller <[e-mail]> wrote:
> If you conclude that the TLS wrapper covers all security concerns that are 
> addressed by MD5, then you can leave MD5 as non-approved. But this then does 
> not hurt because as it does not contribute to security. Thus, you do not 
> violate a FIPS requirement.

  That's good to know.

  It would be useful if NIST published a statement to this effect.  Confusion around the FIPS requirements causes ongoing issues for most RADIUS implementors.

  The customer says "FIPS says no MD5, so we must ban ALL uses of MD5."

  The RADIUS vendor responds with "You can have a completely open network with no authentication, or you can have a secure network using RADIUS with MD5.  Pick one".

  At which point peoples heads explode.  There is much arguing back and forth.

  Corporate processes deal better with checklists than with exceptions.  A simple "FIPS means no MD5" is a checklist item, and is clearly understandable by everyone involved.  It is a lot of work and frustration to explain that RADIUS is somehow special, and that it's fine for RADIUS to apparently violate the FIPS requirements.


  As a similar example, there are many claims hat RADIUS sends passwords "in the clear" over the network.  This isn't true, of course.  But a naive reader sees "clear text passwords" and "RADIUS sends passwords", and leaps to the wrong conclusion.  This issue is made worse by endless "security" experts writing public articles which are 100% wrong.

  This has been a source of confusion and misunderstanding for decades.  I can't tell you how many times I've had to argue with "security" people who are implementing some kind of ridiculous checklist for their network.  They can't seem to grasp that if they're talking to RADIUS people, then perhaps those RADIUS people might just understand RADIUS better than they do.


  The issue of FIPS and MD5 is similar.  Arguably we should have removed MD5 entirely from RADIUS/TLS when RFC 6614 was published.  Since that hasn't happened, we have the choice either to address the issue, or to spend decades explaining why it's fine to use RADIUS in FIPS mode.

  Alan DeKok.