Re: [secdir] SECDIR Review of draft-ietf-radext-design-05

[Alan T DeKok <[e-mail]>]

Hallam-Baker, Phillip wrote:
> The Security Considerations section could do with some additional work
> however.
>  
> The discussion of encryption of attributes is somewhat confusing.
> Mention is made of encryption, followed by mention of MD5 and SHA1.
> While it was common to describe the use of one way functions to
> obfusticate passwords as 'encryption' in the 1980s, this is not current
> terminology and this needs to be explained.

  OK.  We will update the document to clarify this "obfuscation" !=
"encryption".

> Also I would like to see specific mention made of whatever provisions
> are made for message authentication in the protocol, if none, then this
> should also be specified. This is a major concern in what is essentially
> a protocol that supports the authentication/authorization process.

  The protocol supports a Message-Authenticator attribute, which is an
HMAC-MD5 of the packet contents && secret key.  We can add a note on
this to the security section, though it's already discussed in the other
RADIUS documents, too.

> Finally, I would like to see some mention of the use of a secure tunnel
> such as IPSEC and which types of attributes might need superencryption
> within such a tunnel.

  It may be best simply to reference RFC 3579 && RFC 3580, which already
have extensive discussion of these issues.

  Alan DeKok.