[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Isis-wg] Re: Deprecation of AH header from the IPSEC tool kit
>>>>> "RJ" == RJ Atkinson <rja@inet.org> writes:
RJ> At 10:23 19/06/00 , Paul Koning wrote: Nothing in the ESP or AH
RJ> specs prevent the key from being known at an intermediate point.
RJ> So the assumption that the key is only known end-to-end isn't
RJ> always true.
>> I don't know about other people, but I'm not about to even
>> consider the notion of a "security" protocol where the key is
>> known all over the path. No way.
RJ> Paul,
RJ> I didn't say "all over the path". I said "at an intermediate
RJ> point". The distinction is crucial.
RJ> It is quite reasonable to authenticate traffic as it crosses an
RJ> administrative boundary as part of implementing a administrative
RJ> security policy. This is roughly the function of most stateful
RJ> inspection firewalls, though those generally have limited
RJ> assurance because they cannot authenticate the packet headers.
RJ> With AH authentication, a firewall or security gateway can
RJ> achieve much higher assurance than without AH authentication. I
RJ> am aware of folks working on such products.
Fair enough. It's common enough to put a security gateway in the same
box as a firewall. For example, our own Access Point products do that
today. Then again, they do that with ESP, which does all that is
needed for this application.
RJ> Also note that AH can be used with public-key digital signatures,
RJ> rather than the commonly used symmetric hash.
Huh? I see nothing of that in RFC 2402. In any case, given the
performance (or rather, lack thereof) of digital signatures, it's
unlikely that any extensions to AH or ESP to use digital signatures on
a per-packet basis would see any significant use.
paul
Follow-Ups:
References: