Gentoo Archives: gentoo-dev

From: Spider <spider@g.o>
To: gentoo-core@g.o
Cc: gentoo-dev <gentoo-dev@g.o>
Subject: [gentoo-dev] on the matter of security and cryptography
Date: Sun, 04 Aug 2002 11:59:07
Message-Id: 20020804185531.3b8dcf3f.spider@gentoo.org
1 Well, I was browsing around the other day checking into the "behind the
2 scenes" work of gnupg and keyservers when I stumbled on this link..
3
4
5 http://www.cryptnet.net/fdp/crypto/strong_distro.html
6
7
8 Could you please take the time to read this and then consider:
9
10 What would be the requirements to getting this into our current system
11 without breaking backwards compability?
12
13 How do we do it to enforce "minimum developer effort"
14
15 What other infrastructure tools would be necessary?
16
17 and, how would we go about setting up a key-ring for Gentoo linux
18 developers who are completely unable to meet up in large quantities?
19
20 its a difficult concept and hard to do while maintaining security for
21 all involved parties.
22
23
24
25 What I see here as some ideas are.:
26 *) gpg is included as a base dependency for the system and included on
27 default bootdisks...
28
29 *) gpg import is done at "rsync" and the public keyring is avaiable on
30 rsync servers (better ideas people?)
31
32 *) gpg hash/checksum is done on the .ebuild, tarball and patches that
33 are avaiable through portage. (stored in digest)
34
35 *) all developers keys are linked in a keyring.
36
37
38 Now, all this is just stalling, it will be quite difficult to get all
39 the complex aspects of this "safe" into portage, so this will require
40 very very careful design before we start coding. having a rapid
41 development language is no excuse for a crappy design solution in this
42 case.
43
44 Key security is always an issue and a weak point when it comes to
45 developing software. acknowledged.
46
47 how do we avoid infringment into the keys (unauthorized keys added?) and
48 thus enabling an attacker to sign the modified ebuilds/patches and have
49 them check as clean?
50
51 we need to implement hashes for all patches and included files as well,
52 not only for tarballs. since if the patch is applied, it doesn't matter
53 if the tarball was ok, the system is flawed.
54
55 we need to do all this with as little user intervention as possible.
56
57 Now, this is cc'd to -dev as I think it should be in the open, its a
58 proposal and the time for discussion, and a following idea discussion
59 could do well in a larger community. we do have some bright minds, and
60 now is the time for even those who cannot code to come with ideas and
61 insights, because nothing goes into code, just yet.
62
63
64 Regards,
65 Spider
66
67
68
69 --
70 begin .signature
71 This is a .signature virus! Please copy me into your .signature!
72 See Microsoft KB Article Q265230 for more information.
73 end

Replies