gnu-arch-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gnu-arch-users] Re: MD5 is broken


From: Peter Conrad
Subject: Re: [Gnu-arch-users] Re: MD5 is broken
Date: Wed, 16 Mar 2005 12:39:10 +0100
User-agent: KMail/1.6.2

Hi,

Am Mittwoch, 16. März 2005 11:51 schrieb Karel Gardas:
> On Wed, 16 Mar 2005, Peter Conrad wrote:
> > Hi,
> >
> > On Wed, Mar 16, 2005 at 12:26:30PM +0600, Ivan Boldyrev wrote:
> > > Tom Lord merges sexy patch.  Even if he will re-sign patch,
> > > MD5 sum in ./checksum will be same because *.patches.tar.gz is same.
> >
> > this is wrong. If Tom merges your patch, he will automatically create
> > additional log entries in his own branch. This (among other things, like
> > changed timestamps) will lead to a file with a different MD5 sum.
>
> I'm afraid the whole message is a bit different: hack the mirror, hack the
> patch while keeping MD5 intack and let your attack to software X spread
> thorough the world.

I understood Ivan's scenario like this:

1. attacker creates Patch-A (harmless) and Patch-B (evil) with identical
   checksums
2. attacker submits Patch-A to maintainer
3. maintainer integrates Patch-A into software, signing it
4. attacker hacks mirrors and replaces signed Patch-A with Patch-B

To which I answered that step 3 will normally change the MD5 sum that's
actually signed. Which means that replacing the patch will invalidate
the signature.

> I've just now looked at tla and baz and found that at least mirror on:
> http://bazaar.canonical.com/archives/address@hidden/ uses also
> SHA-1 hashes. Since SHA-1 is also considered weak these days, this
> does not add that much security, but certainly at least something
> before arch move to some more secure hash implementation.

Combining different hashes in the signature should make attacks a lot
more difficult, because an attacker would have to produce collisions
for all hashes at the same time. Of course, *all*  hashes must be 
validated when checking the signature, instead of validating only one
of them.

Bye,
        Peter
-- 
Peter Conrad                        Tel: +49 6102 / 80 99 072
[ t]ivano Software GmbH             Fax: +49 6102 / 80 99 071
Bahnhofstr. 18                      http://www.tivano.de/
63263 Neu-Isenburg

Germany




reply via email to

[Prev in Thread] Current Thread [Next in Thread]