Digital Signatures for the official Vim binaries on Windows

[Philip Taron]

Hey all,

I noticed for some time now that the official Vim binaries distributed
on vim.org for Windows users aren't digitally signed.

Is this due to lack of funds, lack of desire, technical limitations,
or personal choice?

If it is lack of funds, I'd like to donate so this could happen.

Philip

[Taylor Hedberg]

I can't answer your question, though I suspect it's just because not
that many people feel that it's important, and a large portion of
"advanced" users (at least those not on Windows) either build Vim from
source or obtain a binary version via their operating system's package
manager, which usually means it's signed by the OS maintainers.

But I'm curious, why would it cost money to do this? GnuPG is free, so
whatever the reason, I doubt that it's a monetary issue.

[Tony Mechelynck]

IIUC, Bram's binaries are (outdated but) signed: see either of the MD5
and MD5SUMS files in the ftp://ftp.vim.org/pub/vim/pc/ directory.

If youwant an up-to-date Vim for Windows, I recommend Steve Hall's "Vim
without Cream", http://sourceforge.net/projects/cream/files/Vim/ � that
one doesn't seem to be signed but is it Steve's or SourceForge's policy?

Best regards,
Tony.
--
God is a comic playing to an audience that's afraid to laugh.

[Philip Taron]

I'm talking about Authenticode signing, where the binary contains signing and repudiation information. There are a couple problems with signing outside the binary:

1. The file itself doesn't contain any provenance information

2. There's no way to verify that the catalog file containing the MD5s is itself signed (or originated from a trusted source)

This leaves alone the problem of using MD5, which is no longer useful for digital signature verification. Wikipedia's got a good writeup on why... http://en.wikipedia.org/wiki/MD5#Security

My main reason for desiring this is summed up here: http://www.hanselman.com/blog/UsingCodeSigningCertificatesToSignDownloadedMSIsAndBuildReputationWithIE9SmartScreen.aspx

Here's examples of the differences between signed and unsigned binaries: http://imgur.com/a/7xJK0 (I used a recently downloaded version of Firefox as an example.)

Cream distro -- well, that one suffers from the same problem. I'd prefer to use the vim.org/Bram build of Vim if I can, since I can be sure it is fully up to date and doesn't have janky personal customizations and patches. 

Why does it take funds? Because not everyone can be a certificate authority. There is a chain of trust that originates in the set of root certificates installed on everyone's machines, and self-signed certs must be manually added on every machine that wants to trust that author is who he or she claims they are.

Philip 

On Mon, Jan 2, 2012 at 5:48 PM, Tony Mechelynck <antoine.m...@gmail.com> wrote:

On 03/01/12 00:11, Philip Taron wrote:

Hey all,

I noticed for some time now that the official Vim binaries distributed
on vim.org for Windows users aren't digitally signed.

Is this due to lack of funds, lack of desire, technical limitations,
or personal choice?

If it is lack of funds, I'd like to donate so this could happen.

Philip

IIUC, Bram's binaries are (outdated but) signed: see either of the MD5 and MD5SUMS files in the ftp://ftp.vim.org/pub/vim/pc/ directory.

If youwant an up-to-date Vim for Windows, I recommend Steve Hall's "Vim without Cream", http://sourceforge.net/projects/cream/files/Vim/ — that one doesn't seem to be signed but is it Steve's or SourceForge's policy?

[Steve Hall]

On Mon, Jan 2, 2012 at 8:48 PM, Tony Mechelynck wrote:
> On 03/01/12 00:11, Philip Taron wrote:
> >
> > I noticed for some time now that the official Vim binaries
> > distributed on vim.org for Windows users aren't digitally signed.
> >
> > Is this due to lack of funds, lack of desire, technical
> > limitations, or personal choice?
>

> IIUC, Bram's binaries are (outdated but) signed: see either of the
> MD5 and MD5SUMS files in the ftp://ftp.vim.org/pub/vim/pc/
> directory.
>

> If you want an up-to-date Vim for Windows, I recommend Steve Hall's
> "Vim without Cream",
> http://sourceforge.net/projects/cream/files/Vim/ — that one doesn't

> seem to be signed but is it Steve's or SourceForge's policy?

No policy, but I'd be curious to know what the OP believes to be
practically accomplished with signed files. Perhaps we're just talking
about the official binaries? Or just checksums?

--
Steve Hall [ digitect dancingpaper com ]

[Ben Fritz]

On Jan 2, 9:41 pm, Steve Hall <digit...@dancingpaper.com> wrote:
>
> No policy, but I'd be curious to know what the OP believes to be
> practically accomplished with signed files. Perhaps we're just talking
> about the official binaries? Or just checksums?
>

Maybe so you can install on Windows without the "are you sure you want
to run this file from an unverifiable source?". If it were signed I
think it would be a less scary message.

Personally I don't really care :-)

[Steve Hall]

On Mon, Jan 2, 2012 at 10:28 PM, Philip Taron <philip...@gmail.com>
wrote:

> I'm talking about Authenticode signing, where the binary contains
> signing and repudiation information.

[...]

> My main reason for desiring this is summed up here:
> http://www.hanselman.com/blog/UsingCodeSigningCertificatesToSignDownloadedMSIsAndBuildReputationWithIE9SmartScreen.aspx
>
> Here's examples of the differences between signed and unsigned
> binaries: http://imgur.com/a/7xJK0 (I used a recently downloaded
> version of Firefox as an example.)

This is a Microsoft scare tactic, there's no reason not to trust
software if you are confident of where you got it. You can eat food
from state certified restaurants and get sick, or eat at a neighbor's
house and feel great. (I'd even argue the latter is safer.)

So I'd love to see the point made using Free Software and not
requiring license fees or key hosting by whatever corporation. (Unless
the case is being made that only state sponsored food should be
allowed.)

> Cream distro -- well, that one suffers from the same problem. I'd
> prefer to use the vim.org/Bram build of Vim if I can, since I can be
> sure it is fully up to date and doesn't have janky personal
> customizations and patches.

You obviously don't get the point of Free Software. :)

> Why does it take funds? Because not everyone can be a certificate
> authority. There is a chain of trust that originates in the set of
> root certificates installed on everyone's machines, and self-signed
> certs must be manually added on every machine that wants to trust
> that author is who he or she claims they are.

It only takes funds because the crooks that are trying to scare
everyone into a fully sponsored "security solutions" need money to
survive.

[Philip Taron]

No policy, but I'd be curious to know what the OP believes to be
practically accomplished with signed files. Perhaps we're just talking
about the official binaries? Or just checksums?

I'm only interested in the official binaries. The problem of determining that the sources retrieved from the official master repository are the same sources is something else entirely.

1. Integrity. I know the binary has not been modified in transit in some form. Catalog signing, like the MD5 file talked about here, also accomplishes this, provided that there is something that signs it, and so on.

2. Identity. I know that the person claiming to be Bram Moolenaar (or Steve Hall, or whomever) is certified to be that person by some certification authority I already trust.

3. Authorship. Combining the previous benefits, I know the file is intact, that Bram really is a person/org, and that he really produced this file. 

4. Provenance. I know that the binary I got from vim.org actually originated with someone who both controlled vim.org and also the private cert for codesigning the binaries there. (This is only if vim.org supports https, which it currently does not.)

5. UX benefits. I'm restricting this to Windows, since I have no idea of the state of PKI/code signing/etc on Linux or MacOS. On Windows, executables that are digitally signed are presented differently than binaries which are unsigned.

6. Revokability. If the prior constraints do not hold true (due to a systems failure, vulnerability, or loss of private key, for instance) the certificate can be revoked immediately.

7. Individual revokability. If a particular binary suffers from a very bad vulnerability, it can explicitly be pulled.

There's also a couple white-listing benefits, which are completely ancillary.

8. Anti-malware benefits. Most AV engines (and in particular the one used by MS, for instance in Security Essentials) are able to author whitelist signatures for known good certs.  

9. Reputation services (like Smart Screen for downloads in IE9). Over time, these can provide actual trust benefits (like http://www.hanselman.com/blog/UsingCodeSigningCertificatesToSignDownloadedMSIsAndBuildReputationWithIE9SmartScreen.aspx illustrates.)

In reality, my personal motivation is to get rid of that damn unsigned dialog, but from an objective standpoint my motivations don't matter. :)

Philip

[tux.]

No one said vim.org will never be compromised.

[Philip Taron]

 

This is a Microsoft scare tactic, there's no reason not to trust
software if you are confident of where you got it. You can eat food
from state certified restaurants and get sick, or eat at a neighbor's
house and feel great. (I'd even argue the latter is safer.)

Dare I note that both sourceforge.net and vim.org are not offered over https? Without that, there's no way to know whether I'm eating at a mockup of my neighbor's house or at the house itself.

 

So I'd love to see the point made using Free Software and not
requiring license fees or key hosting by whatever corporation. (Unless
the case is being made that only state sponsored food should be
allowed.)

> Cream distro -- well, that one suffers from the same problem. I'd
> prefer to use the vim.org/Bram build of Vim if I can, since I can be
> sure it is fully up to date and doesn't have janky personal
> customizations and patches.

You obviously don't get the point of Free Software. :)

Hey, enough with the hate, suffixed with smiley faces as it is. Anything prefaced with the phase "I prefer" surely is meant only in a personal manner. More power to you for creating and maintaining Cream. It's not _my_ preference.

 

> Why does it take funds? Because not everyone can be a certificate
> authority. There is a chain of trust that originates in the set of
> root certificates installed on everyone's machines, and self-signed
> certs must be manually added on every machine that wants to trust
> that author is who he or she claims they are.

It only takes funds because the crooks that are trying to scare
everyone into a fully sponsored "security solutions" need money to
survive.

Root of trust, distribution of keys, revocation, and the other associated issues with a global PKI are real problems. In a free software context, see the hack on kernel.org and GNU savannah...

http://blog.sucuri.net/2010/11/savannah-gnu-org-hacked-and-currently-offline.html

https://www.linuxfoundation.org/news-media/blogs/browse/2011/08/cracking-kernelorg  

Digitally signing the binaries wouldn't have eliminated either of these problems, but would have made cleaning up after them quite a bit easier.

Philip 

[Philip Taron]

> Cream distro -- well, that one suffers from the same problem. I'd
> prefer to use the vim.org/Bram build of Vim if I can, since I can be
> sure it is fully up to date and doesn't have janky personal
> customizations and patches.

You obviously don't get the point of Free Software. :)

Hey, enough with the hate, suffixed with smiley faces as it is. Anything prefaced with the phase "I prefer" surely is meant only in a personal manner. More power to you for creating and maintaining Cream. It's not _my_ preference.

On consideration, I apologize for the "janky" characterization. It was uncalled for.

Philip

[Ernie Rael]

On 1/2/2012 8:43 PM, Philip Taron wrote:

Dare I note that both sourceforge.net and vim.org are not offered over https? Without that, there's no way to know whether I'm eating at a mockup of my neighbor's house or at the house itself.

 

When I log into sf.net I start getting an https URL.

-ernie

[Bram Moolenaar]

Philip Taron wrote:

It's a lot of hassle to get this certification, costs quite a bit of
money (several thousand dollars), and only gives a little bit of
protection. The obvious way around it is to just replace the signed
binary with a not signed binary, hardly anyone would notice.

In practice messing with the files has never happened and if it did it
would most likely be detected and fixed quickly.

Trojan horses are a big problem, but the signature is a very weak
protection against them.

--
If cars evolved at the same rate as computers have, they'd cost five euro,
run for a year on a couple of liters of petrol, and explode once a day.

/// Bram Moolenaar -- Br...@Moolenaar.net -- http://www.Moolenaar.net \\\
/// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\
\\\ an exciting new programming language -- http://www.Zimbu.org ///
\\\ help me help AIDS victims -- http://ICCF-Holland.org ///

[Philip Taron]

It's a lot of hassle to get this certification, costs quite a bit of

money (several thousand dollars), and only gives a little bit of
protection.  The obvious way around it is to just replace the signed
binary with a not signed binary, hardly anyone would notice.

In practice messing with the files has never happened and if it did it
would most likely be detected and fixed quickly.

Trojan horses are a big problem, but the signature is a very weak
protection against them.

I'll drop the topic. Thanks for providing the current consensus opinion.

Philip