James Gray wrote:
.
You're incredibly trusting for a guy who believes that the only secure kernel
is one you compile yourself. You're happy to trust your mirror's/distro's
user-land apps, but not the kernel? WTF?
(http://lists.slug.org.au/archives/slug/2004/11/msg00081.html)
The context of my proposal and practice to recompile kernels is to
enable me to include
only the modules that I needed and exclude those that I do not need to
miminize the
opportunities for attacks and to minimize maintenance. This is for
high-security kernels.
What if I hijack someone's DNS (like planetmirror or pacific internet or
aarnet) then start spewing trojanised packages all lovingly signed with my
"fake" GPG-KEY? The only way you'd know my packages weren't kosher was if
you had the ORIGINAL key from the original package source.
Again check the idea of one-way-hashing or message-digest. It is
precisely the maliciousness of hijacking that it is meant to prevent.
I know what what I one-way-hash is, I did "Computer Science 101" too. Your
idea only works as long as I'm checking all the packages against the
published key from the original packager. If I sign the packages with a
malicious key, and you download the malicious key, then you'll be unaware the
packages have been altered.
Hijacking, as you probably know, is someone in the middle of two
participants in communication that pretends to be one-of-them and changes
the messages between thes two participants.
You didn't read my entire message did you? I'm not talking about a
man-in-the-middle attack. I'm talking about hijacking someone's DNS and/or
BECOMING the original mirror. Not intercepting traffic to/from the mirror.
That's why you have to have GPG-key to ensure that the file is not changed
whilst in transit. One-way-hashing would prevent this situation.
Only if you get the original hash (or signing key) for the original package.
If I trojanise a package, create a new hash/key based my own "malicious" key
and package, then you'd be happily downloading my trojans blissfully unaware
of the fact, until you get the original signing key from the original source.
Remember, I'm talking about a malicious mirror here, set up for the purpose of
spewing trojanised/back-doored packages. *NOT* a man-in-the-middle attack.
If it were the latter, then your comments would be true. My comments (and
Glen's) are based on the same principle for why you don't sign untrusted
GPG/PGP signatures. Get the sig from the source, always.
Cheers,
James
--
SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/
Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
