6 ms·

Verifying Wikileaks DKIM-Signatures

76 points by teknotus 10y ago

dooglius 10y ago
I didn't realize the signatures were leaked as well. Surprised nobody has looked at this yet, if we have cryptographic proof that certain emails are genuine, that shoots down a big counter-argument of "what if some emails are planted".
- gregschlom 10y ago
  > if we have cryptographic proof that certain emails are genuine, that shoots down a big counter-argument of "what if some emails are planted". Huh, no? I don't see how that follows logically. If we have cryptographic proof that all email are genuine, then yes. Right now we already know that some (most?) email are genuine. The question is whether or not the attacker, who was obviously motivated by political purposes, also inserted some fakes in the lot. (Although to me that seems highly unlikely)
  - dooglius 10y ago
    Good point, but we know specifically which emails are genuine, and which may or may not be. Any particular email that isn't signed is questionable, but we can say for any given email whether or not it might be a fake.
- koolba 10y ago
  If you click the "View Source" section, they have the entire MIME message. The display you see is formatted to make things easier to browse and list out any attachments (if any).
dzdt 10y ago
Yes, the interesting thing would be if any emails with controversial content can be verified. I don't think there were any real bombshells in the wikileaks emails, but still...
- thomask0 10y ago
  Most of the things i've came across that are being brought up as bombshells are actually incoming mails from outside the campaign. Plus there is a number of things that are being put out of context like the Hillary hates Everyday Americans thing, where it's clear from the context that it is just about the phrasing
  - gravypod 10y ago
    There are a lot of things that are horrible in these leaks. For example, here are four: http://www.usatoday.com/story/news/politics/onpolitics/2016/10/13/four-juiciest-leaked-podesta-emails/92014368/ http://www.usatoday.com/story/news/politics/onpolitics/2016/... From those four, the following are "verified" by the OP of this thread. - 5205: The Clinton's seem to have had access to the questions to a debate. - 4178: The Clinton's seem to have been advised about the investigation into their emails from someone in the DOJ. These are two out of 4. I'll edit this if someone links other articles and I'll check them across this link.
    - thomask0 10y ago
      The DOJ status hearing is one of the perfect examples of something out of WikiLeaks that was way blown out of proportion by tweeting first, ask questions later. http://mediamatters.org/research/2016/10/11/nbc-reporter-suggests-scandal-over-routine-status-hearing-regarding-release-clinton-emails/213754 http://mediamatters.org/research/2016/10/11/nbc-reporter-sug... http://www.politico.com/blogs/under-the-radar/2016/10/trump-clinton-wikileaks-doj-229615 http://www.politico.com/blogs/under-the-radar/2016/10/trump-... On the debate thing, a Bernie senior aide came out and said that she gave guidance to Sanders campaign too. The most nefarious scenario would be that Brazile gave questions to Clinton to prop her up. The less nefarious scenario would be that Brazile from her DNC position tried to get the best out of both candidates to not embarrass the DNC. Note that Brazile and CNN are still denying she could have gotten access to the townhall questions. I agree though that networks should not put themselves in these kind of conflict of interests, by hiring a political commentator that is paid by the DNC at the same time, or a former campaign manager that is still on the payroll of said campaign like Corey Lewandowski. http://www.latimes.com/nation/politics/trailguide/la-na-trailguide-updates-former-senior-aide-to-bernie-sanders-1476297181-htmlstory.html http://www.latimes.com/nation/politics/trailguide/la-na-trai...
xentronium 10y ago
I remember having issues verifying emails with attachments last time I've used dkimverify.
- teknotus 10y ago
  Any insight into what breaks verification is useful. Thanks!
koolba 10y ago
Very cool research! Is there any proof-of-existence archive of DKIM keys or DNS record changes in general? Seems like a perfect use case for blockchain[1]. If the DKIM keys were rolled over or replaced since they were originally sent out, there wouldn't be anything to compare them against. Having a record that can show proof-of-existence (at a min point in time) would cover that. [1]: Ha! I knew there would eventually be a legit use case!
- robryk 10y ago
  This just requires the inability-to-rewind property of the blockchain, doesn't require there to be one centralized log of events and doesn't require protections against flooding with data (if we verify that the DNS records are correct at the point of inserting). This makes a certificate-transparency-like solution just as good and much less complicated (one can arguably call that blockchain, but usually people mean by that things that can support a currency).
  - EdHominem 10y ago
    Well, we've been making validated structures where a checksum (essentially a pre-cryptographic hash) of the last data-block is included in the next since the sixties. (And I suspect much further back in an accounting context.) The difference these days is a protection, via proof-of-work, against just creating a new, valid, chain by editing the old one. That proof of work, to be translatable into a dollar value, requires a market for the proofs and that's the currency part. (Without that you don't know the true market value of the energy required for the PoW.)
Torgo 10y ago
There has been no evidence yet that the emails were doctored. What there has been, as one journalist put it, was efforts to conflate "wikileaks emails" with "unsourced shit I saw on Twitter".
- thomask0 10y ago
  No evidence that they were doctored, but there are still plenty of complete fakes that were tweeted out by others, which were taken serious by way too many people including FOX news. http://www.snopes.com/hillary-calls-voters-bucket-of-losers/ http://www.snopes.com/hillary-calls-voters-bucket-of-losers/
source99 10y ago
Some quick numbers for those interested: Excel sheet has 11104 rows 2423 verified - 22% 8681 not verified - 78%
pbarnes_1 10y ago
DKIM only signs headers. This is bunk. Move along.
- agwa 10y ago
  You're wrong. As you can see on page 31 of RFC6376, the signature clearly covers a hash of the message body: https://tools.ietf.org/html/rfc6376#page-31 https://tools.ietf.org/html/rfc6376#page-31 It's possible to hash only a prefix of the body by specifying the 'l' parameter in the DKIM header field, but this is unusual and the email used as example in this blog post does not use the 'l' parameter.
StoneSolone 10y ago
I think the major problem of this election is deciding on what the truth is but since everything can be rewritten, links broken, and data lost there is too much noise to get a clear signal. We need a trust/identity protocol that allows us to take history in account I wrote a draft of a whitepaper on a possible way to implement this with blockchains and cross-signing https://hashd.in/hashd-in-draft0/ https://hashd.in/hashd-in-draft0/

76 points by teknotus 10y ago

I didn't realize the signatures were leaked as well. Surprised nobody has looked at this yet, if we have cryptographic proof that certain emails are genuine, that shoots down a big counter-argument of "what if some emails are planted".

gregschlom 10y ago
> if we have cryptographic proof that certain emails are genuine, that shoots down a big counter-argument of "what if some emails are planted". Huh, no? I don't see how that follows logically. If we have cryptographic proof that all email are genuine, then yes. Right now we already know that some (most?) email are genuine. The question is whether or not the attacker, who was obviously motivated by political purposes, also inserted some fakes in the lot. (Although to me that seems highly unlikely)
- dooglius 10y ago
  Good point, but we know specifically which emails are genuine, and which may or may not be. Any particular email that isn't signed is questionable, but we can say for any given email whether or not it might be a fake.
koolba 10y ago
If you click the "View Source" section, they have the entire MIME message. The display you see is formatted to make things easier to browse and list out any attachments (if any).

dzdt 10y ago

Yes, the interesting thing would be if any emails with controversial content can be verified. I don't think there were any real bombshells in the wikileaks emails, but still...

thomask0 10y ago
Most of the things i've came across that are being brought up as bombshells are actually incoming mails from outside the campaign. Plus there is a number of things that are being put out of context like the Hillary hates Everyday Americans thing, where it's clear from the context that it is just about the phrasing
- gravypod 10y ago
  There are a lot of things that are horrible in these leaks. For example, here are four: http://www.usatoday.com/story/news/politics/onpolitics/2016/10/13/four-juiciest-leaked-podesta-emails/92014368/ http://www.usatoday.com/story/news/politics/onpolitics/2016/... From those four, the following are "verified" by the OP of this thread. - 5205: The Clinton's seem to have had access to the questions to a debate. - 4178: The Clinton's seem to have been advised about the investigation into their emails from someone in the DOJ. These are two out of 4. I'll edit this if someone links other articles and I'll check them across this link.
  - thomask0 10y ago
    The DOJ status hearing is one of the perfect examples of something out of WikiLeaks that was way blown out of proportion by tweeting first, ask questions later. http://mediamatters.org/research/2016/10/11/nbc-reporter-suggests-scandal-over-routine-status-hearing-regarding-release-clinton-emails/213754 http://mediamatters.org/research/2016/10/11/nbc-reporter-sug... http://www.politico.com/blogs/under-the-radar/2016/10/trump-clinton-wikileaks-doj-229615 http://www.politico.com/blogs/under-the-radar/2016/10/trump-... On the debate thing, a Bernie senior aide came out and said that she gave guidance to Sanders campaign too. The most nefarious scenario would be that Brazile gave questions to Clinton to prop her up. The less nefarious scenario would be that Brazile from her DNC position tried to get the best out of both candidates to not embarrass the DNC. Note that Brazile and CNN are still denying she could have gotten access to the townhall questions. I agree though that networks should not put themselves in these kind of conflict of interests, by hiring a political commentator that is paid by the DNC at the same time, or a former campaign manager that is still on the payroll of said campaign like Corey Lewandowski. http://www.latimes.com/nation/politics/trailguide/la-na-trailguide-updates-former-senior-aide-to-bernie-sanders-1476297181-htmlstory.html http://www.latimes.com/nation/politics/trailguide/la-na-trai...

xentronium 10y ago

I remember having issues verifying emails with attachments last time I've used dkimverify.

teknotus 10y ago
Any insight into what breaks verification is useful. Thanks!

koolba 10y ago

Very cool research! Is there any proof-of-existence archive of DKIM keys or DNS record changes in general? Seems like a perfect use case for blockchain[1]. If the DKIM keys were rolled over or replaced since they were originally sent out, there wouldn't be anything to compare them against. Having a record that can show proof-of-existence (at a min point in time) would cover that. [1]: Ha! I knew there would eventually be a legit use case!

robryk 10y ago
This just requires the inability-to-rewind property of the blockchain, doesn't require there to be one centralized log of events and doesn't require protections against flooding with data (if we verify that the DNS records are correct at the point of inserting). This makes a certificate-transparency-like solution just as good and much less complicated (one can arguably call that blockchain, but usually people mean by that things that can support a currency).
- EdHominem 10y ago
  Well, we've been making validated structures where a checksum (essentially a pre-cryptographic hash) of the last data-block is included in the next since the sixties. (And I suspect much further back in an accounting context.) The difference these days is a protection, via proof-of-work, against just creating a new, valid, chain by editing the old one. That proof of work, to be translatable into a dollar value, requires a market for the proofs and that's the currency part. (Without that you don't know the true market value of the energy required for the PoW.)

Torgo 10y ago

There has been no evidence yet that the emails were doctored. What there has been, as one journalist put it, was efforts to conflate "wikileaks emails" with "unsourced shit I saw on Twitter".

thomask0 10y ago
No evidence that they were doctored, but there are still plenty of complete fakes that were tweeted out by others, which were taken serious by way too many people including FOX news. http://www.snopes.com/hillary-calls-voters-bucket-of-losers/ http://www.snopes.com/hillary-calls-voters-bucket-of-losers/

source99 10y ago

Some quick numbers for those interested: Excel sheet has 11104 rows 2423 verified - 22% 8681 not verified - 78%

pbarnes_1 10y ago

DKIM only signs headers. This is bunk. Move along.

agwa 10y ago
You're wrong. As you can see on page 31 of RFC6376, the signature clearly covers a hash of the message body: https://tools.ietf.org/html/rfc6376#page-31 https://tools.ietf.org/html/rfc6376#page-31 It's possible to hash only a prefix of the body by specifying the 'l' parameter in the DKIM header field, but this is unusual and the email used as example in this blog post does not use the 'l' parameter.

StoneSolone 10y ago

I think the major problem of this election is deciding on what the truth is but since everything can be rewritten, links broken, and data lost there is too much noise to get a clear signal. We need a trust/identity protocol that allows us to take history in account I wrote a draft of a whitepaper on a possible way to implement this with blockchains and cross-signing https://hashd.in/hashd-in-draft0/ https://hashd.in/hashd-in-draft0/