Secrecy and user trust

[Bill Davidsen]

Patrick O'Callaghan wrote:
> On Thu, 2008-09-04 at 23:12 +0800, Ed Greshko wrote:
>> Patrick O'Callaghan wrote:
>>>> NAK - if a fake public key were distributed then packages signed with 
>>>> the fake key would be matched, allowing full access to install crap in 
>>>> your machine.
>>>>     
>>> True.
>>>   
>> Actually I don't understand the paragraph above.  It seems to be saying
>> that packages would be signed with a public key which can't be done. 
>> So, the person making that statement needs to clarify.
> 
> Which is the point I made earlier.
>   
>>>> And packages signed with any valid redhat key would be 
>>>> rejected.
>>>>     
>>> Which is what I said. Thus it would be noticed immediately.
>>>   
>> No, they would not be rejected as long as you still have Red Hat's
>> public key installed on your system.  You can determine what public keys
>> are on your system by "rpm -qa gpg-pubkey*". 
>>
>> When an rpm is signed it is signed with a private key and information
>> about the corresponding public key is placed in the rpm file.  That
>> information is used to retrieve the correct public key for
>> verification.  So, as long as you've not deleted it, they will verify.
> 
> The hypothetical scenario being discussed is that you have already
> replaced the former (good but now possibly suspect) public key with a
> spurious new one. If that were to happen, you would be in danger of
> accepting trojanned packages signed with this new fake key. My point is
> that you would also *reject* packages signed with the new good key, and
> this would be noticed very quickly (basically the next time you did an
> update).
> 
That's exactly right, and why the public key should be as trustworthy as 
possible, because once you accept a single trojanned package you may 
either suffer damage immediately, or have some part of the "next time 
you did an update" fail. Imagine a slight change to updated so it never 
tells you there *is* an update.

I am making the point that an improved method of checking the new key is 
desirable, technically possible, and that a false package could cause 
problems in a very short time, and might be able to hide thereafter.

-- 
Bill Davidsen <davidsen at tmr.com>
   "We have more to fear from the bungling of the incompetent than from
the machinations of the wicked."  - from Slashdot