[Cryptography] DNSSEC = completely unnecessary?

[Lodewijk andré de la porte]

2013/11/6 Paul Wouters <paul at cypherpunks.ca>

> I've answered Moxie (and others) on this before:
>
> https://nohats.ca/wordpress/blog/2012/04/09/you-cant-p2p-
> the-dns-and-have-it-too/
>
> ExecSum: you cannot avoid trust, making it hierarchical gives the least
> trust to parties. You monitor those you have to trust more, more
> closely.
>

A simply dumb article. "America" refers to a place in the Netherlands and
also that huge continent. There's a certain decentralization in the fact
that my nation picked that name for a place and the USA can't do anything
but ask us to change the name. They didn't because nobody is actually that
confused.

How? Context.

So what's the solution that removes the need for single source of Truth?
One to Many relationships in the domain name system. Then some "context"
mechanisms build into your local DNS daemon and browser to deal with the
ambiguity. I'd imagine a simple "vote up or down" would suffice to give the
right domain name to the right people. Provided you, at least practically,
solved the identity problem. Possibly though a Web Of Trust.

I can also imagine central source being plugged in as a "According to
ICANN:". Kind of like "sponsored results" in Google.

The idea that a method for agreeing upon a name with a larger group of
people requires structure is valid. Language itself proves that. In fact
language is a beautiful analogy.

In other words, address agreements are not “peer to peer”. There is a
> centralized, or rather a hierarchical method for assigning, reassigning and
> distributing location name updates. And the only reason it works, is
> because there is censorship and central control.


Yeah, they are. Words' meanings are P2P too. The thing is you often allow a
peer to make choices, a government for example, or simply comply with the
most common agreement (a surprisingly common protocol). Censorship is a
completely unnatural idea. The natural alternative is filtering, and it is
done at the peer level. That's a form of self censorship and it remains a
problem, imagine a community not propagating .xxx domain names. Anyway.

You do not need central control. You need an agreeable and compatible
protocol. You do not need censorship, you need ranking.

Simply ranking ICANN 100% trusted  and everyone else 0% is not only
simplistic but also foolish. You seem to want to find out why.


> The Domain Name System is a lookup mechanism similar to the Geo Location
> Name System (AKA “atlas”)

Which atlas? You'll note there's quite some differences, and there have
been. But DNS is mostly consistent (every day!).

The DNS serves to match an easy to remember mnemonic (“nohats.ca”)

Unless the easy one was taken. Also: say that to all the non-ascii users!
Took quite a while before naver.com could even become 나볼.한. In fact there
isn't even a .한! Go figure!


> to an impossible to remember IPV4 or IPv6 address.

That's a bit of a big statement ain't it?

> It ensures that no two entities can claim the same mnemonic.

Unless some nation (cough china cough) would decide they'd rather have
their own authority. I'd call this the biggest flaw in DNS history.

> And it is also really really big. If we look just at the COM, NET and ORG
> we’re talking in the order of magnitude of 100 million entries.

There go all the easy to remember names. See also "parked domains". You
love those don't you? Just imagine a ranking system wiping this form of
spam to irrelevance!


> If you do not live in Canada, your atlas likely does not contain the exact
> location of 100 Queen Street West, Toronto, Canada. If it did, your house
> would be filled with just your atlas.

It's called a computer. It can contain such information in less volume than
your fingernail. Welcome to real life.

So let’s say you have the perfect decentralized P2P DNS table. It can
> uniquely represent every “domain name”,

Internal contradiction or empty statement. The most preferred page can be
said the unique representation, but you map away a lot of information.

Reading your page I just can't figure how you think this problem is so
hard! It's really *really *simple!

Let’s assume Verisign had to give the NSA copies of the private key of the
> root zone, and the NSA can sign anything from the root down.

This is so obviously the case. Wow. I mean just wow. The rest of the
article ignores knowing that things like TLS and, well, everything else is
also MITM'ed. If you're lucky something will complain to you, and if it
does it can mean a whole ton of things. Besides, you're probably not lucky.

And let's not talk about the buckets of unused OS level exploits the NSA
has. I mean what are we talking about here.

"We had this map of the EFF's SSL Observatory data on what countries are
> currently capable of intercepting secure
>
>>       communication under the CA system. Under [DNSSEC/DANE], it would
>> look like this."
>> https://www.youtube.com/watch?v=Z7Wl2FW2TcA#t=33m43s
>>
>> (He shows a completely red map, indicating all countries)
>>
>
> That's a nonsense argument. Abuse of such powers, unlike the plethora of
> CA certs, would need to be world visible, that is untargetted. It would
> be very very visible. It is a huge win over CAcerts that can target
> individuals with specifically crafted signed certs.
>

MITM. I am the master of what you perceive to be a reality. Argument
invalid.


> With dnssec, if the Government of Canada causes my nohats.ca to be
> modified (appear red on your above map), then my domain's public
> information changes. I would notice that. This is not an invisible
> MITM like some CA cert injection.
>

Well, yeah, and the server's SSL connection would also give you this
information. HTTPS makes DNSSEC redundant is not exactly an awesome counter
argument. Especially because DNSSEC does not make HTTPS redundant.


> Which ever organisation you pick for the initial "name and crypto key
> lookup" has that power. What you must do is track and monitor, so ensure
> that power is not abused.
>

Design flaw: I'm an end user. I don't even know what crypto is. (Monitor...
Wasn't that the thing I'm looking at?)


> It's very easy to say "trust no one". But if you want to talk to people,
> you need to place trust in some. All replacements I have seen of
> DNS(SEC) just move the problem elsewhere.
>

The most robust of them all. Good old selfish-incentives.

 You want to have a hierarchical trust pyramid.
>
No.

> You can monitor, and even safeguard by picking some parts on the pyramic
> you trust more than the top (root).
>
Which application lets me do this on my tablet again? What about my desktop
pc? What about my fucking wifi-enabled printer? Will these trust profiles
sync automatically? Or does it just not matter if my e-faxes are mailed to
a different domain by my printer? Oh I shouldn't use my printer.

Thanks for nothing.

> This both scales and keeps the trust needed to be
> given to a minimum.
>
Except that those I trust I give my life to.

> The higher in the hierarchy a trust organisation is, the less targetted
> violation of trust they can do. Moving down, trust is only handed down to
> entities that can only betray themselves, not others.
>

Sounds false. I am really bothered by discussing something based on wrong
axioms. You probably don't deserve my given amount of hate. Sorry about
this. But really, you expect too much from users, intermediates and
software especially. SSL has had trouble. DNSSEC will have trouble too.
It's not magical.


> Any kind of "harvest and vouch" or "public ledger" solution is going to
> be riddled with false positives, due to the delay between
> publisher/owner of the data and the trustee/ledger updating itself.
> Plus you don't even _know_ which parties it is that you are given
> trust to when accessing "N of M" entities to determine truth.
>

Yeah I don't really know what you meant by that.

DNSSEC is nice in that if you trust the root (or something closer by in the
chain) and the records are signed than you have a place to get an SSL
public key. That way you have two checkpoints, the server and the SSL chain
and the DNSSEC chain.

In practice I feel there'll be no great advantages because the signing
giants will be the same giants. Mostly American (wonder where they got
their initial funding!).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20131107/3c04e9b6/attachment.html>