On 9/2/15, Paul Vixie <p...@redbarn.org> wrote: > > > John R Levine wrote: >> ... >> >> Tor is one approach to query security that seems to work pretty well >> give or take side channel leakage. Dunno if there are any others, but >> it is clearly a very hard problem, and not one we're going to solve >> any time soon. > > i think we have to be realistic. the national security apparatus of the > largest dozen economies have effectively unlimited resources. they will > spend as many of our tax dollars as it takes to buy as much equipment > and talent and methodology and electric power as it takes to watch what > they decide is important to watch.
What of the national security apparatus of the rest of the countries of the world? Or the criminal cops buying HackingTeam software? Or perhaps a teenager with tcpdump? Until the invention of quantum computers, we can protect data from being instantly available to most of these groups most of the time. Right now, we're not doing any such thing and in fact it is the opposite: we're failing to protect data from all of these groups, all of the time. I do agree though - we have to watch out for people that they try to buy. There is easy no technical solution to dealing with people who stop protections from being deployed. We know the NSA sends people to the IETF and specifically to make protocols easier to intercept and exploit. One might guess that other services do similar things. How might we find such people? How might we ensure that they stop harming users? I think we won't and so rather we should finding flaws which benefit such an adversary and fix them. > > i'm all for PFS on the cache miss path Could you identify all of the parties involved, what they do and which ones in your view should have privacy? It sounds like you want to support certain kinds of end user privacy. I'm extremely curious to hear a specific proposal of who deserves and who doesn't deserve privacy. My outline is as follows: everyone and every system should have security and privacy in the form of forward-secret authenticated cryptography, all of the time, enabled by default. Ideally we should design the link level protocols to look identical and to ensure that selector based surveillance is ineffective for any data relating to DNS inside of the encrypted protocol. Thus I expect a sniffer may see that it is encrypted DNS traffic but will be limited to that as an XKeyscore AppID. The others selectors available will be IP, port and other metadata leaked by the TCP/IP protocol itself. > but trying to hide who the query > came from so that an authority server has less insight, is nutball > stuff. I disagree and strongly. We need protocols that compose with other protocols. This needs to be one of the design goals. It won't happen over night. I already have this property when I use Tor with some other software - shouldn't everyone else have it too? > remember that the NSA (et al) is a form of "us" and that we will > not stop "us" with math. real change there will come in the voting booth. With the utmost respect, I believe your theory of change lacks evidence and flies in the face of what we've seen in practice. We've found that the math is the main thing that changes the economics of mass surveillance. Power, space and cooling are the other major factors. Agencies that tap fiber already *break* laws or the laws are written to ensure that non-US persons aren't considered as people. While I support a limited view of your theory of change in small pockets of the world, I believe that no matter how you or I vote - the fact remains: the BND, the PLA or the FSB won't care about your or my American vote, likewise with nearly every other country in the world we'll get what technology ensures for us. We need to step up our game and the zeroth step is to recognize that we need to change the technical and the political situation at the same time. Deploying forward secret tech is what is currently lacking for most protocols. In DNS the situation is worse. There is political will in many countries around the world but change stops when the technology isn't up to the task. Pervasive monitoring as discussed in RFC 7258 is an attack. DNS currently fails on a technical level to address this kind of attack. All the best, Jacob _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
