| 1 |
* Paul de Vrieze (pauldv@g.o) wrote: |
| 2 |
> Date: Wed, 24 Mar 2004 11:36:32 +0100 |
| 3 |
> From: Paul de Vrieze <pauldv@g.o> |
| 4 |
> To: gentoo-dev@l.g.o |
| 5 |
> User-Agent: KMail/1.6.1 |
| 6 |
> X-Spam-Status: No, hits=0.0 required=5.0 tests=none autolearn=no version=2.63 |
| 7 |
> Subject: Re: [gentoo-dev] 2004.1 will not include a secure portage. |
| 8 |
> |
| 9 |
> -----BEGIN PGP SIGNED MESSAGE----- |
| 10 |
> Hash: SHA1 |
| 11 |
> |
| 12 |
> On Wednesday 24 March 2004 00:36, Jesse Nelson wrote: |
| 13 |
> > so a dev key would be revoked by a sync on a file containing a list |
| 14 |
> > of dev keys ? what if i compromise a rsync server adn a developers |
| 15 |
> > box. This isnt that far fetched a scenario. I get acess to a dev key |
| 16 |
> > and i stop a server from updateing the signed key sigs file. |
| 17 |
> |
| 18 |
> That's why the signing key needs to be shortlived. Basically if the |
| 19 |
> signing key is valid for 1 day, the list of dev keys becomes invalid |
| 20 |
> after a day. That would halt the rsync part of the problem. Of course it |
| 21 |
> would still be a compromise and we would need to check all files signed |
| 22 |
> by the key, but it would be needed anyway as it is unlikely we can stop |
| 23 |
> the intrusion at the moment it happens. |
| 24 |
|
| 25 |
not bout stopping intrusion just that its a verry likely poisioning scenario, and even if the signing key dies after 1 day.. if i got root on the server i just update. besides now theres 1 days worth of sync's out in the wild with compromised builds/pataches/binaries. |
| 26 |
> |
| 27 |
> > this also doesnt begin to solve the rogue developer problem (tho thats |
| 28 |
> > not the goal on this i guess) |
| 29 |
> |
| 30 |
> There is no way to stop this before that person is identified in any |
| 31 |
> case. After this person is identified his keys will be revoked and all |
| 32 |
> the packages signed by him/her are invalid. They will need to be |
| 33 |
> resigned by someone else to be valid again. |
| 34 |
|
| 35 |
you can stop this buy: |
| 36 |
having multiple eyes have to see the changes b4 a commit to rysnc servers, and it follows that you would then have multiple sigs for a item (build/dist/patch) etc or multiple sigs on a sums file. |
| 37 |
|
| 38 |
this could also go through a security review b4 goin out live. Instead of say relying on ppl cross-checking. A defined security team could have to sign b4 release, but i doubt eaither of these would go over well with already overloaded devs. |
| 39 |
> |
| 40 |
> > imho every package needs 2 pairs of eyes b4 gettin released. (2 dev |
| 41 |
> > sigs + sign key) but theres procedural and time constraints that will |
| 42 |
> > probly prevent that from ever happening. |
| 43 |
> |
| 44 |
> That is a QA issue, but would not change the basic infrastructure. I |
| 45 |
> think we should first discuss single signatures and then only discuss |
| 46 |
> multiple signatures. |
| 47 |
|
| 48 |
its a QA issue.. its really a process issue.. building QA/security into the release cycle. |
| 49 |
|
| 50 |
> Paul |
| 51 |
> |
| 52 |
> - -- |
| 53 |
> Paul de Vrieze |
| 54 |
> Gentoo Developer |
| 55 |
> Mail: pauldv@g.o |
| 56 |
> Homepage: http://www.devrieze.net |
| 57 |
> -----BEGIN PGP SIGNATURE----- |
| 58 |
> Version: GnuPG v1.2.4 (GNU/Linux) |
| 59 |
> |
| 60 |
> iD8DBQFAYWS9bKx5DBjWFdsRAr7cAKDdO71+8LiHg9KR2E1pZp7QyqAVvQCgrOXY |
| 61 |
> EPRjTlCzStpL5DNQCKh8T8U= |
| 62 |
> =INcS |
| 63 |
> -----END PGP SIGNATURE----- |
| 64 |
> |
| 65 |
> -- |
| 66 |
> gentoo-dev@g.o mailing list |
| 67 |
> |
| 68 |
> |
| 69 |
|
| 70 |
-- |
| 71 |
gentoo-dev@g.o mailing list |
Archives