*** Disclosure: The following is the output resulting from transcribing an audio file into a word/text document. Although the transcription is largely accurate, in some cases may be incomplete or inaccurate due to inaudible passages and grammatical corrections. It is posted as an aid to the original audio file, but should not be treated as an authoritative record.*** DNSSEC for everybody. A beginner's guide. 24 October 2011. ICANN Meeting - Dakar (No audio at beginning). >> -- in the DNS. It works something like this. A resolver is the ISP or the companies. A resolver. Okay. Then -- sorry, your ISP or your company's resolver knows where the root servers are. It goes there, and it traverses (inaudible) How it, basically, works it will hand you the addresses of the com server or the U.K. server. And each level refers you down until you are where you need to be. And the resolver catches all that information for future use. Remember this slide from Simon's deck? This is where Ugwina talks to Og. And Ugwina, in this case, is the resolver that checks with Og. And Og is, basically, an authoritative server, may be responsible for bank.com or for BBC.UK or NIC.SN. In fact, Ugwina talks to many Ogs. So now we're going to take a snapshot of the name space. I just want you to focus today on root com and bigbank.com. And we're going to show you how it works with a little bit of role play. Thanks, Julie. I was scared a bit. So I'm going to quickly introduce you to the players. I'm going to quickly introduce to the players. So on the far left is Joe User. Then we have the root server. Then we have the com server. And then we have the bigbank.com server. And Joe right over here wants to go to www.bigbank.com. >> Thank you. I'm going to do my banking, so I'm sitting at my home computer. I'll type in www.bigbank.com. >>ROY ARENDS: I am the ISP's resolver. I don't have that information cached. I do have the information for the root server, so I will go there first. Root server, I want the address for www.bigbank.com. >> Hi, Mr. ISP. I haven't seen you for a while. I'm afraid I can't help you with www.bigbank.com, but I can tell you where dot come is. It's at 1.1.1.1. >>ROY ARENDS: Next up dot come at 1.1.1.1. Hello, I want the address for www.bigbank.com. >> Hello, ISP. Good to see you again. So I don't know the address for www.bigbank.com. But I do know that you can find it at 2.2.2.2. >>ROY ARENDS: Thank you. So, allegedly, the address for www.bigbank.com is available from bigbank.com. Hello, bigbank.com. I want the address for www.bigbank.com. >> Hello, Mr. ISP. I can happily give you www.bigbank.com since I am bigbank.com. Www.bigbank.com is at 2.2.2.3. >>ROY ARENDS: Thank you. So I now know, as a resolver, where www.bigbank.com is. It's at 2.2.2.3. I'm going to hand that over to Joe User. But not only that, but I'm going to cache that information for future use. >>Thank you, Mr. ISP. Now I can go off do my banking in my browser and pay all my bills. >>ROY ARENDS: Thanks, guys. So what is the issue here? This is actually not a secure transaction what we do. You saw that little piece of paper that we handed back and forth? In a DNS transaction between Ugwina and Og, between the resolver and the server, there's only a small bit more information. But all that information can be spoofed. So in DNS there is no security, no real security. Names are easily spoofed. And, when that happens, those spoofed names, those poisoned names are easily cached. So, when that happens, Ugwina is, basically, confused. Remember that slide again? This is a slide from Simon's deck. She doesn't know who the real Og is. It could be Kaminsky or the real Og. So how does that spoofing work in real life? Guys, I need your help again. So we do the same play here. Joe User. >> Okay. Paying more bills, going to my bank again. So sit down, browser, and type in www.bigbank.com. >>ROY ARENDS: Thank you. I'm going to cut this a little bit short. So I'm first going to talk to the root server. Root server, bigbank -- www.bigbank.com, please. >> Okay. I don't know where bigbank.com is, but I can give you dot com. It's at 1.1.1.1. >>ROY ARENDS: Thank you. 1.1.1.1, I want the address for www.bigbank.com. >> Thank you, ISP. I don't know the exact address of www.bigbank.com, but I do know you can find it at 2.2.2.2. >>ROY ARENDS: Thank you. 2.2.2.2, I want the address for www.bigbank.com. Thank you, bigbank.com. The address for www.bigbank.com is 6.6.6.6. And I know no better than that is the real address I want to go to. Hello, Joe. I have the address for www.bigbank.com. And it's at 6.6.6.6. >> Oh, thank you, Mr. ISP. Now I can go on my computer and do all my online banking. >>ROY ARENDS: So how do we secure this properly? We are going to introduce something called digital signatures. And DNSSEC uses digital signatures to make sure that the answer is received from the right place and has not been tampered with in transit. It uses digital signatures and digital keys. And the good thing about DNS it's a lookup system. You can store anything in the DNS. You can store text records. If you want to send e-mail, that mail transfer agent actually looks up in the DNS. It's a simple lookup system. And so you can store signatures as well in the DNS. And the beauty of this system is that you just use DNS to get the right information. DNS keys are also stored in the DNS. And there's actually a lot more things; but it goes into really, really deep technical detail. If you're interested in that, come see me afterwards and I'll completely bore you with all the details. So remember the blue smoke. That's the important lesson from the first slide deck. Remember the blue smoke. So Ugwina, the resolver, is now trying to verify -- sorry -- is now trying to talk to the real Og. The real Og was actually able to introduce something in his message, blue smoke. Or, in this case, he can show that he used the right key. And everyone can check that key because DNS is open and available. Everyone can use that key to check if the information is correct. So that's not completely it. There's one thing more. A resolver -- remember that a resolver knows where the root servers are. Now, in this case, it actually knows what the root key is. And, just like delegations -- right? -- that builds the hierarchy, you can now have a chain of trust. And that's through that same hierarchy. So, instead of delegations, you now have the chain of trust. In this case we're going to secure -- we're going to make that chain of trust right now. So the root over here at one point in time has learned the key from the com zone. And they shake hands over it. And that, basically, means that he now has a hint for what the proper key is in dot com. Same thing over here. Dot com is able to trust bigbank.com after some initial handshake. And now com has the information to get the key for bigbank.com and to trust the bigbank.com DNSkey. So, again, we now do the high-level concept of the DNSSEC. We do that with the same role play. We start all over again. Go ahead. >> More banking. Type in www.bigbank.com. >>ROY ARENDS: Haven't you learned by now that the Internet is not secure? And broke, okay. So I know what the root server's addresses are. 0.0.0.0. Please don't use it at home. That's not the real address. This is just play. And I also have the key, the public key to trust the information that he has signed. So, root server, I want the address for www.bigbank.com. >> I'm afraid I don't have the address for bigbank.com, but I do know where dot com is. It's at 1.1.1.1, and I'm going to seal that with a handshake. >>ROY ARENDS: Thank you. Because we did the handshake, I can now trust -- sorry -- I trust the root server. I trust everything they sign right now. And the information that I just got from the root server, I can now validate. That's why we did the handshake. Hello, com, I want the address for www.bigbank.com. >> Hello, ISP. Well, I don't have the address for bigbank.com, but I know that you can find it -- for www.bigbank.com, but I know you can find it at bigbank.com. And I'm going to shake your hand to certify that. >>ROY ARENDS: Thank you. I now have the proper address for bigbank.com. I can check it. I can validate it. Okay. That means I now go to bigbank.com. Hello, bigbank.com. I want the address for www.bigbank.com. Thank you. And I'm now going to validate that information. The handshake doesn't work, right? I can't validate the information. I'm going to try again. Hello, bigbank.com. I want the address for www.bigbank.com. >> Hello, www.bigbank.com is at 2.2.2.3, and it is signed and has blue smoke. >>ROY ARENDS: Perfect. This is the chain of trust that I now validated throughout the delegated space, right? And Joe User has no chance -- I'm sorry. [Applause] >> Joe User has a chance to go to the right space, and Dr. Evil has no chance to interfere with that data. Thank you. Okay, guys, that's it. Thanks. So Russ is going to introduce you to some implementation stuff. Deployment guide, basically. Norm, go ahead. >>RUSS MUNDY: Well, I hope everybody enjoyed our little skit. We certainly enjoyed doing it for folks. [Applause] We truly want this to be interactive, and so far nobody has asked any questions. But this part of the session we do intend and hope that we get questions, inputs, and people stop us, raise their hand, whatever. We have another mic. We'll get you out to the audience. But we're going to start off talking about -- let me get the clicker here. Make sure I get it the right way. So I guess I'm -- I'll just go, since that didn't seem to be a question for what we're doing here. Okay. So the focus that we are trying to get to here is a few more details, not huge technical details, but some to show a little bit more of real world examples of how you can go about doing DNSSEC for wherever you might be in the realm of DNS. And so there's generally a set of activities that are owners of DNS zones and names. And then there is users. And these are the people that can be literally anywhere in the world. And so we'll have a little bit of information about each of these, the idea being to give a very broad coverage. So those that are owners of DNS today, owners of names, have a set of responsibilities that they have. They have a set of organizations that they -- (dropped audio) Most people have registrars that they deal with. There are a lot of registrars here in the audience -- or at least here at the meeting. And I hope we have them in the audience. And registrars have a important role in this as well in many spaces. So there's also large enterprises, corporations that have worldwide interests. Some have local interests. Some are just very small activities that have names and Web sites and operate something on the Internet. And they're going to do very different things to actually do DNSSEC, depending on what their role is in the world. And there's a set of other classifications on there that I won't go into specifically. If I can get the advance here. Did I get it? All right. Here's a larger picture similar to what you saw before, but it does give a little bit broader idea. There's -- well, not quite 300 names in the top level root zone. And a small number of them are shown there. And there are literally millions of other names that come down in this tree. We'll be talking today specifically about CNN.com and HP.com, just as a couple known examples. Most people are aware of those two entities. And they're very common on the Internet. So for each activity, wherever you may fit in this structure -- if you're an activity that has an extremely heavy reliance on DNS for the well-being and operation of your business or your function, then chances are you will have a staff and a set of activities, whether it's outsourced or in-house, that's quite knowledgeable in DNS. And, in that case, there's a good chance that you will want to have that set of activities do the things that need to be done to do DNSSEC, whether you're running recursive name servers, authoritative name servers or both. If you're an activity whose DNS is important, but it's not really key to your business and you have everything outsourced and is operated by somebody else, then there's a very good chance that you will have a choice of how you can do DNSSEC. But your most logical choice will be to go to your current DNS provider and say, "I want you to do DNSSEC for me." Now, this little picture is an illustration of the whole content picture for DNS or the zone data picture. The zone data or the DNS content starts clear on the left side of this picture with the registrants, someone getting a name assigned to them. Therefore, they become the owner of that particular name. They deal with a registrar, in most instances, to get that name and to maintain the ownership. And a registrar deals with a registry. That is actually, normally, where the name servers are operated, although they can be separate. And that's clear at the point of the triangle. But the actual content, DNS zone content is used clear over on the right-hand side. So you can see there's authoritative name servers where you were seeing some of us up here playing authoritative name servers earlier. And Roy, as the recursive name server, was in between. And Norm, as the Joe User, was clear over on the other side. You can see the content starts one place and goes way through lots and lots of things to get to somewhere else. Now, another way of looking at this is someone wants to put a www record into DNS. And that starts in this picture clear over to the left in the zone data. It's added into the zone data. It goes into the authoritative name server. And, when it's in the authoritative name server, then recursive name servers and users can get it. This is another way of illustrating what we were showing earlier where Norm, as the client rear on the right, goes through Roy as recursive name server, goes through several authoritative name servers, and gets the answer returned. Another larger picture of what you'd have to do to get www.CNN.com, the root name server constellation is shown at the top. There are 13 different ones. You can go to any one of those. The dot com name servers -- there's 13 at least IP addresses at both the root and com. There are many more actual machines. And then the CNN.com has four separate IP addresses. And then the recursive resolver actually gets that to the person that wants the information. So you look at a web page and say, "Gee, there's not a lot of DNSs there. The URL bar at the top, there's one or two lookups." This is a picture of how many lookups there are about five years ago on CNN.com. And it keeps getting more and more. This is how many there are about three months ago on CNN.com. Well, over 100 queries and responses to get CNN.com web page filled on your browser when you go there. What is all of this? Well, there's advertisements; there's pictures; there's multiple places that the CNN.com web page uses. It's a little hard to see here. The green -- there are a few green lines on the right-hand side. Those are actually DNSSEC validated sites. There's a very small number that you actually do go through when you're filling your web browser with CNN.com. But not that many. So it's the DNS zone data that really matters. DNSSEC gives you the ability as a user to make sure you get the right zone data. But you also have to make sure that the proper DNS zone data gets put in and managed properly in the first place. So it's the DNS zone content that really matters. Now, when you start to look at how you're going to go about doing DNSSEC, some people and some organizations get very carried away because there's crypto involved in DNSSEC. And they have to do extreme things with the security of the crypto. And they don't think enough about whether or not their content has as good of security as their -- the slide numbers? Ah, okay. Thank you. I didn't realize it wasn't up there. So I guess for the remote presentation we don't have the slides. So I'll cover a little bit more of the content. So, when you're assigning your zone as the owner of zone and applying DNSSEC information, you want to also make sure that you put as much effort into the security to make sure that your content that you're putting into the zone has as much security as the cryptographic mechanisms that you're using to sign it. Because, if someone wants to attack your zone and attack applications that are using your zone, if you spend way more resources on the DNSSEC crypto mechanisms and not much resources on the DNS zone content protection, they're going to just attack your content. And then you, as the owner of the zone, will be signing incorrect and invalid data, which some people would argue is worse than not signing a zone and having bad data published. So, if you remember the picture that I had up a little bit ago, the left-hand side is sometimes called the provisioning side. That's the side where you go to as an acquirer of a name, go to a registrar, get approval. And, whether it's exchanging money or your authorization, whatever has to happen to get the authority to use a name, then on the left-hand side of the triangle, the registrar then passes that to the registry. And in the registry area, and from the registry, which is the point of the triangle, until the user application on the far right, that is the area where DNSSEC is actually applied. So, as you can see, if you think of it as a mountain or a triangle, DNSSEC protects you and protects your information on the right-hand side. The left-hand side is no more or less secure than it was before after you have signed it. And, if someone is going to attack your content provisioning part, DNSSEC does not prevent that. It doesn't help it. Doesn't hurt it. It makes it no different. Where DNSSEC really gives you the protection is as the data is going back and forth and passing all around the Internet. And, if you'll recall, there are many hundreds of places that it goes through. And a pretty typical flow of a query and a response will go from anywhere to 10 to maybe as much as 30 or 40 different nodes. And all along the way on the Internet flow of things is where attacks can occur. With DNSSEC in place, those attacks are thwarted. Again, back to the earlier picture, the picture where you're putting your World Wide Web additional content, the record to tell people in the world where your World Wide Web server is into your authoritative name server, now with DNSSEC, when that information goes into your authoritative name server, it will be signed information. And we'll have the blue smoke attached that can be used by the validating recursive resolver or at some point even an end user, if validation occurs there. And so the part of the Internet where literally hundreds of packets flow around for a given single query to fill CNN.com, you're protected. Now, some more specifics about where and how people should look at applying DNSSEC. And, as I said earlier, the general principle is, whatever you're doing for your DNS operation today, you should probably do in the similar or a same manner for DNSSEC. So, if you're running your own DNS operation -- you have your own operation staff; you have your own computing systems; you have your own internal operation and, essentially, you're running your own ISP, then you're probably going to have that same staff apply DNSSEC capabilities into your existing infrastructure. If you're going to do that, if you're using a given defined product suite, then you will likely want to use the product suite's DNSSEC implementation. And now, when you go to the product suite vendor and they say, "Oh, I'm sorry. We don't do DNSSEC," that's when it becomes your part to say, "You need to do DNSSEC, because I want it and I want it as soon as I can get it." And you might even think about changing vendors to get it. And be sure to tell your vendor that. That gets them motivated pretty well. Now, if you're using open source tools in your internal operation, there's a number of choices you have. There are some appliance type operations -- available that are open source. There are appliances that are products that can be easily added, integrated with an open source operation. There are also signing services that you can use in conjunction with your open source operation. They are very easily integrated into your operation. Certainly, I'm not going to go into a lot more detail there. It's available. If anybody has any questions, say so. We'll talk about it now. Or we can cover it afterwards, if you like. If you're using an outsource kind of operation, then you'll want to go to that outsource provider and say, again, "I am using you today for DNS. I want to use you for DNSSEC." And, hopefully, their answer will come back, "Oh, certainly. You just look at this particular box on the Web site where you interface to me and click it off, and you will have DNSSEC." If that's not the answer, if the answer is something like, "Oh, I don't offer that. That's nothing that anybody would really want to do" or something of that nature, you say, "Well, I'm sorry. That's when I need to start looking for somebody else to provide my DNS service that can provide DNSSEC for me." And truly, this is a very, very important aspect. Because this is the kind of motivation that, whether it's outsource providers, whether it's product providers, any other entity that is providing a service related to DNS, if they hear from their customers that they have customers wanting DNS security, DNSSEC, they will be more likely to do it. And, when they start losing customers if they don't, then they'll definitely get motivated to do it. Now, there are a range of specific products that I've not tried to cover any here. And we do conduct an informal survey that's available on the DNSSEC-deployment.org Web site that will give a breakout of a large number of these. And we can provide more specifics there. But it seemed that I didn't want to really cover any individual ones here. And I wanted to just go through these alternatives and give people the opportunity to see whether -- where they fit into things. Now, personally, I have a soft spot in my heart for all the open source things because, frankly, that's what we do a lot of. And we have many things available that are free and openly available in the space. And there are a number of other products that are also open source. But many people do use commercial products, and the availability from commercial product vendors is not as much as we would like to see. And, truly, what we really need is to have people who are customers of products asking their product vendors to provide DNSSEC. So now I want to see if there are particular questions that folks might have relative to anything that they have seen today. We have got another mic, so we can pass it around. And back in the back, Julie, there is somebody back this. And it can be anybody that you have heard from. We'll take any questions that you have. >> I have a simple question. In the drama that you showed, the bogus data was trying to mimic the valid IP. What would happen if the bogus IP latched onto it. Is that possible? Latched on to trusted, signed data. >>RUSS MUNDY: Okay. I'm not sure that I got the question. Could the wrong ISP? Was that the question? >> Basically, Russ, I think the question is can somebody inject bogus data at an earlier phase so that it's returned with a trusted signature? >> Yes, exactly. >>RUSS MUNDY: Okay; good. We must have more questions than that. We do. Now, I thought Jay gave -- I'm sorry. >> (Off microphone). >>RUSS MUNDY: I am having a hard time hearing in this room. >> It is possible, yes, for the data, through a breach of the registrant or through the registrar, which has happened in a number of occasions so that somebody has taken control of somebody else's Web site through breaking into the registrar they use and then sending the bad data that way. And theoretically it is also possible to break into a registry and insert the data in the registries systems so that by the time it makes it into the DNS system, it has the associated signature with it. So it's a known vulnerability. >> (Off microphone). >> It has happened, and it has happened a number of times in the last two years with registrars being compromised and Web sites being redirected for high-profile targets as a result of that. So the threat to DNS that DNSSEC is protecting is far more widespread and far more exploited than the threat of people breaking into registries. >>RUSS MUNDY: Thank you, Jay. Because I lost it in the overhead sound. Sorry. Go back to the triangle picture in the slides. I think that's the easiest place to point to. Right there. The registrant/registrar area that over here on the left-hand side of this picture is the area that I think you were asking about, that information really just affects the -- if it's a registrar attack, it affects only the zones handled by that particular registrar and doesn't go all over the Internet. The side that DNSSEC is protecting, once it gets on the Internet, this is extremely broad. It goes anywhere and everywhere and could be any zones on the Internet. So the protection is required on both sides but the damage that can occur is broader and wider on the side that DNSSEC is protecting. More questions. Back here. >> Thank you. The way I understood it is that DNSSEC is basically an authentication function that's being integrated into DNS. What is the advantage of doing it this way rather than implementing this function at the level of the Web service? >>RUSS MUNDY: Okay. The difference is that DNS is used by literally hundreds probably thousands of applications throughout the Internet. If you were attempting to solve the same problem universally and as broadly as DNS can solve it, you would have to solve it for those hundreds to thousands, maybe even hundreds of thousands of different applications. And each solution would probably be different, where solving this and ensuring that people go to the right name is usable by any and every application that runs on the Internet. Now, they can do additional security mechanisms on top of it that are useful for their particular application functionality, but they can focus on that application functionality for their security and not have to worry about whether or not the name that they get is correct. Is that -- >> Yeah, the thing I was thinking about was that this will probably take a while before DNSSEC is being rolled out on a very large scale. So maybe there is some sort of intermediate phase where you still have to rely on the security from the Web sites. >>RUSS MUNDY: Well, a Web site is one example. And that's probably the largest, most broad spread set of examples. But there has also been a growth in DNS security that has been very effective the last few years. There are now over half of the domain names that are registered can operate with DNSSEC; okay? With the signing of the large TLDs as we have. And so I don't know the exact numbers. I know it's well over half, though, with com and net, DE, UK, and the other large top-level domains. So it's going -- it's being rolled out quicker than a lot of people thought. Is there one over this? Yes, Roy. >>ROY ARENDS: Not a question but an addition to what Russ has said. Before you can actually do the Web stuff, you need to get there first, and you use DNS for that. So DNS is the only place you can really secure that. So to make sure the information is correct, you go to the proper place, you need to have DNSSEC. >> (Off microphone). >>ROY ARENDS: Hold on, hold on. I thought it was simple. >> So if I am doing online banking today, then I am -- and even the banking site is secure, then I'm still not sure if it can be a bogus site. >>ROY ARENDS: If you do banking today without DNSSEC, you might have a certificate, DigiNotar, you might have a certificate that could be secure, could not be secure. Users are mostly presented with the certificates that they then have to trust. People are building now whole applications -- for instance, things like as DANE, or DNSSEC stable certificates, so you can put certificates in the DNS so you have the proper certificates before you actually do the banking, before you go to that Web site. >>RUSS MUNDY: One other -- I think that gentleman is next, but one other thing. There have also been a couple of high visibility CA compromises. There's also a number of other questionable CAs that get distributed -- the root certificates get distributed in Web browsers that a user wouldn't even get a warning for even if they are, in fact, compromised. So it is a different situation, but it's also recently one that's been showed to be attacked with great rigor and some success. >> Are there any additional fees to have DNSSEC work with my registered domain name? If I go to a registrar and I register a new domain name and I want to use DNSSEC, is there an additional fee to do that or is it included as part of the service, typically? >>RUSS MUNDY: The costing of DNSSEC is completely up to those that are issuing the names. Okay? So it varies by where you are in the world and what the organization is that you are dealing with. >>JULIE HEDLUND: We have some over here, too. >> Can you comment on what the expected market adoption is likely to be from a time perspective? And then within that, are there any segments such as e-commerce companies that might be earlier adopters than others? I'm curious as to what you see in the marketplace. >>RUSS MUNDY: A very important question, the adoption rate. It's also been very hard to answer. We have looked at this several times over time. We have had some market segments that, especially the financial, we thought would be early adopters. They have not been. There is some movement in various countries in particular to start to use DNSSEC within the financial industry. The area that we have really seen the broadest pick up and adoption has been on the authoritative and top-level domain side to make sure that the service gets out there, gets available, so that it is broadly available. And right now, the most visible example that I'm aware of of a large company that is into the financial world that is publicly committed to doing DNSSEC is PayPal, and they have also said in some public statements that they intend to require PayPal partners to also do all DNSSEC. >>JULIE HEDLUND: And, Russ, we are really getting some good questions here and so we are going to need to start managing a queue. So I have seen we have got one, two, we have got three. We have this gentleman over here, we have this gentleman here. So I am going to start to this side and move to this side, and we will -- >>RUSS MUNDY: You are in charge of the queue, Julie. >> Thank you. I came in late so I don't know if this question was already talked about. I'm more worried about the security. So you can have a good name, a religious name or anything like that. Sometimes you type that, somebody can have it as a dot com, somebody else as a dot FR or dot org or something like that. Now you type that name and you get redirected to an adult site. You following me so far? >>RUSS MUNDY: No. >> Okay. So I'm asking you about the effectiveness of the DNSSEC here when it comes to security. >>ROY ARENDS: So I think I understand the question. Let me repeat that for you, if that's okay with you. So the question is really how does it protect against, for instance, things like typos or redirection. >> Exactly. >>ROY ARENDS: And the answer, it doesn't. If you type in the wrong name, you go to that wrong name. Actually, DNSSEC makes sure you go to that wrong name. In DNSSEC terms, us geeks, we sometimes say, you know, bank robbers, they wear seat belts as well, where bank robbers are basically those who try to steal traffic away from the right person and try to rob you from your money; right? And the seat belts is basically the DNSSEC. So it doesn't help on the provisioning side at all. It doesn't help, it doesn't do anything for you in terms of redirection. Part of the question could also be how does DNSSEC help against, for instance, influence of governments or things like COICA or mandated DNS redirection. DNSSEC only helps so little. It can only detect that data has been tampered with. From a DNSSEC perspective, if an ISP decides to reroute data or just send you to an alternative address, to DNSSEC that means the data has been tampered with and it will tell you that. It doesn't mean you will get, then, to the correct data. It probably means you don't get this at all. Thank you. >> Sometimes even with the seat belt, you can die. >>ROY ARENDS: Yes. >> Okay. >> I am a student and I want to use the ICANN as a reference. So my question is to ask you if you deliver certificate at the end of the seminar. Thanks. >>RUSS MUNDY: Okay. I'm going to, I think, turn this over to Roy. I am just having too hard of a time hearing. >>ROY ARENDS: Okay. What we can do, we have some -- I think you are asking for reference information where we can find information about how to deploy DNSSEC; is that correct? >> She is actually asking if you guys deliver certificates for her. She is a student. She wants to use it as a reference, and if you guys could actually put her with a mentor or something like that. >>ROY ARENDS: Okay. Sure. We can do that. Come see me afterwards. Thank you. >>JULIE HEDLUND: Excuse me, but we do have a queue, and I have noted this gentleman here but we have this gentleman and the one over this. And then that gentleman. >> Excuse me. I speak French but I don't understand English very much. (Speaking in French) (Scribes have no translation) >>ROY ARENDS: I am going to answer you in English. I'm sorry. The first question, for those who haven't heard the translation, I am going to shorten your question into two or three key sentences. Your first question, Joe User does a bank transaction. Joe User sees a certificate being presented to him from a bank. DNSSEC gives him additional assurance. How does the experience for Joe User work? What information does Joe User have to assert that DNSSEC has been done, basically? I am waiting till the translation is done. Is that correct? >> (Speaking in French) (Scribes have no English translation) >>ROY ARENDS: Okay. What's the difference between SSL, on one hand, and DNSSEC on the other hand. I read that question as, for instance, a bank can currently issue -- authenticate the data stream. Joe User can see the lock. Everything is secured by SSL. What does DNSSEC do? DNSSEC makes sure that -- let me turn that around. Before you actually get the certificate from that bank, you have to know where that bank is. So that's one. Now, you might have gone to the -- You have might have gone to the right bank. You might have gone to the wrong place. And what you see a lot is that people are presented with a certificate, and something is wrong there, and you can click on yes, no, maybe. What these people -- and it includes everyone, including my mom, including myself. What you want to do, because you want to did your banking, you just want to get past that obstacle. So you are just going to click on the certificate and accept it and take a leap of faith. Now, of course, that's not smart, but we know from studies that certificates alone is not secure enough. We also know, for instance, events like DigiNotar or COMODO, they have been hacked. Now, of course that can happen in DNSSEC as well, but it is the assurance you go to the right place. So before I actually do the banking, I can have some assurance that I went to the right place. Currently DNS validation happens mostly at the validator -- sorry at the ISP level. What we would really like to have eventually is that validation happens in the application or on the operating system, so that the end user can see both that the transaction is SSL secured and they ended up in the right place. Now, for your second question, your second question, in order to, let's say, play with DNSSEC and in order to test it, is it possible to have a separate independent environment, basically a lab system or compartmentalized system. I can answer that in two ways. You can do that, but you need to remember that just like -- you basically need to mimic everything. You need to compartmentalize everything. Just like you can have root servers, the proper root servers that are currently run by ICANN, IANA, and VeriSign, you can have your own root servers, basically a small set of servers. So that way you mimic the entire namespace and you can play there on your own test with DNSSEC. That's one answer. If you want to use the regular namespace, the namespace we are all using -- for instance, in Senegal, dot SM -- what you can do is instead of configuring in your resolver the root DNSkey, you can configure the, let's say, playground.academic.something.SM, if you will. So it depends on how you configure it but it is definitely possible, and if you would like to have some examples, we can chat afterwards. We need a translator. I don't understand French. But I will help you with some examples. Thank you. >>RUSS MUNDY: While we are going to the next person, let me just comment that this was a study done at the Massachusetts Institute of Technology about three or four years ago where there were 100 students from MIT that were told that their transaction with their bank, their interaction with their bank was going to be mucked with in some manner, and they weren't told how. And of those 100 students, 93 of them accepted something and said okay to something they should have never said okay to, and, therefore, would have compromised their banking activity. So it's not to say that having security in an application is not necessary or is bad in any way. It's just that people have become so used to just accepting these things that don't look right to them that the strength, the protection is not as much as it used to be or was originally. >>SIMON McCALLA: We have a question from the chat room, if that's okay. Jay has posted a question. He says: My question would be to know if the kind of warranty that DNSSEC offers to me is in any way more secure or equivalent to a pin generator and a VPN connection that my bank provides me with? >>RUSS MUNDY: I would say it's very much orthogonal to what a bank would offer, because there is no standard universal strength or rigor associated with that. It could be extraordinarily strong. It could be very weak. And you just don't have a way, nor do I, of assessing the strength or weakness. The DNSSEC mechanisms are all defined by public specifications. Therefore, you can at least judge how strong you believe the result of that is. >>SIMON McCALLA: I think the key to this one is to remember, and I think something that Roy mentioned is that DNS is the pathway to get to your bank in the first place. And the key thing about DNS is this a security standard built into the very fabric of the Internet rather than additional tools and technologies that are applied on top. That is why we believe DNSSEC is so important and can sit alongside additional mechanisms provided by your bank. So it's an additional layer of security, too. >>RUSS MUNDY: Yes. >>AKSHAT: Hi, this is Akshat. I just wanted to ask, before your schedule mentioned a term called cache poisoning. Does DNSSEC cater to that in any way? >>RUSS MUNDY: Cache poisoning was actually the original biggest threat that DNSSEC was created to counter. At the time that DNSSEC engineering first got under way, it was extraordinarily easy to do cache poisoning, much easier even than it is today, and that was the original biggest threat that it was put out there to prevent. And it's one of the reasons why, in the engineering of DNSSEC, the security is part as an integral part of the data of the DNS zone itself, as opposed to having something along the paths or trying to secure every individual name server. That was the biggest reason for that choice to be made, to put it with the DNS data itself. So no matter where it went, if there was something that went wrong in a cache, the eventual user of it could detect that. >> Okay. Thank you. >> Domain name security seems like a wonderful idea, and it seems to be very old, at least 5,000 B.C. So why have we only been looking at it for the last three years? What suddenly was the kick up the rear end? >>RUSS MUNDY: You want to take that, Roy? Okay. >>ROY ARENDS: As Russ said before, cache poisoning is almost as old as the Internet, is almost as old as the DNSSEC. What happened in the last three years is cache poisoning got incredibly easy. Really incredibly. I remember an IETF -- For those of you who do not know what IETF meetings are, that's basically environment where standards are created that everyone can use on the Internet, and we had a meeting in Ireland in 2007 -- sorry, in 2008 or early 2009, where we were discussing if DNSSEC would be the answer to the Kaminsky attack, to the cache poisoning attack. And people did not grasp that cache poisoning was so simple. And a former colleague of mine, he went on stage and he gave a small demo that in less than a second, he was able to spoof information to poison the cache. It was a very simple trick that he did. It was very, very fast. And still people didn't believe him, and he did it again and again and again. So this was 2008. And folks are still not convinced that cache poisoning is that simple. But at that time, during the IETF, it was basically the equivalent of shooting a gun in a room where people were talking about gun control. It was very, very impressive. So. So the 2008 Kaminsky attack really made this thing go forward. But DNSSEC has been around since I think 1995. The first -- When people first understood that there was something wrong, it was, I think, Steven Bellovin came up with a method to poison the cache, and folks started thinking about how to make that more secure. So that's basically the history of DNSSEC, the real history of DNSSEC. Does that answer your question? Okay. Thanks for asking. >> Thank you. Hello? Yeah. My question is this. Based on the demonstration, you have four actors. In the initial scenario -- is it five? The ISP inclusive with the root, and the second-level domain and the third-level domain; right? The actual owners. That's the registrant. Are they all supposed to be signed? If the root is signed, we are talking about -- I am from a ccTLD. The ccTLD is signed. Do all the registrants, the end users, need to sign their domains as well? Or is the signing of the registry solve -- takes care of that? That's the first question. The second question is this: How does an ISP uses DNSSEC? Because we are talking about the ISP DNS server; right? That you use to browse the Internet. The third question is how as an end user, how do I validate whether the ISP is using DNSSEC or not? >>RUSS MUNDY: Okay. So the first question, do you need to have all the domains in the hierarchy signed for DNSSEC to be effective. For the owner and operator of each zone in question, they need to have their zone signed. And so if, in the case of our little skit here, if bigbank.com was not signed, and the user really wanted to get to the host in bigbank.com, he could not get DNSSEC validation only by having the root and com signed. All of the levels have to be signed. Now, realistically in the DNS world today, there's a small amount of interactions with, if you will, the general public and TLDs, but most of the interactions are in the level below the TLDs. So, yes, you do need to get to that third level of having DNSSEC signed zones for users to be able to conduct -- use DNSSEC effectively. Now, you will know getting so far but you won't know clear to the end host. So that's the first question. The second question in terms of how does an ISP use it or provide it, the ISP in our skit was Roy; okay. Roy was acting as the recursive name server for an ISP. Now, there's -- really, you can think of DNSSEC and DNS as having two basic parts. One basic part is the building of the DNS zone content information itself, the zone data. That's the creation process or, on that slide, it's the left side, the provisioning side. As part of that, and right at the peak of that triangle, is where DNSSEC is normally applied. So that has to be put into the system. Once that's put into the system, anyone operating a recursive name server that has the root key can validate down the chain. So that's how an ISP would actually use it. They would operate a DNSSEC validating recursive resolver. So was that -- >> That question was the end user. How do I know whether my ISP is using that or not? >>RUSS MUNDY: How would an end user know? Today, you would have to ask. There are a small number of applications available that do do DNSSEC on the end system. I have got some of them on my machine that we build and distribute. But the vast majority of the Internet for a long time will probably be using recursive resolvers provided by their ISPs. Now, one of the things that we have been working with through some of our projects is with Mozilla and other Web browsers, open-source Web browsers, to contribute code that does DNSSEC validation and get it distributed through these large major open source browsers. So when that occurs and it becomes readily available and widely available, it will be a much easier process for an end user to actually do DNSSEC on their end machine. >> Thank you very much for the explanation. The last question, please. I am a paranoid end user. I really want to secure my domain, but my ccTLD is not signed. Is it possible to sign my domain? >>RUSS MUNDY: You certainly can sign your domain. Now, if you are Roy's example a little earlier to the gentleman that was asking can you set up a separate secure subdomain, you can if you have some reasonably small number of people that you are dealing with that you want to make sure that they know that it's coming from your domain. You can provide out-of-band secured handing of your public DNSSEC key to those activities. But it will only work for those activities that you have handed your key to, and whenever you change it, you also have to make sure that those activities have the changed material. So you can do it. >>ROY ARENDS: My answer is a little bit shorter. You can use a thing called DLV. If your top-level domain is not signed, there is a trick you can use in DNSSEC currently, is use the concept of DLV. It basically means that you side step your top-level domain. So let's talk about it afterwards. I can help you with the content. >>JULIE HEDLUND: I think it was over here. Was there someone over on this side, hand up? I've got it. >> Hi. Thank you for your presentation. I would like to know for the DLV, are they able to play the same role of root signed zone? >>RUSS MUNDY: Another quick DLV question; right? Is it able to play the same role? Is that your -- >> Because as I understand DLV, we use it when the root zone or the top-level domain zone is not signed. So I would like to know, is it able to replace the root zone signed? >>RUSS MUNDY: Well, the root zone is signed, many of the TLDs are signed. But if you happen to be in a TLD that's not signed, then you can use the DLV, DNSSEC lookup site validation to, as Roy said, step around the lack of signing in the zone right above your zone. So the DLV has been running for about five years. Is that right, Norm? I think. Yeah, about five years. So it's an established activity out there and will continue length of time until there is a lower-down hierarchy number of zones that get signed. So it's not exactly the same as having a fully signed path, but as long as the users of your zone data are willing to do validation with the DLV, you functionally get the equivalent of a signed tree. >> I have another question. It's about the keys. How -- With the trust anchor, the keys for signing the zone are managed by DNSSEC. The keys allowing to sign the zone are managed by DNSSEC. >>RUSS MUNDY: Well, the actual policies and mechanisms for handling of the keys are determined by each zone owner operator, just as any of the DNS information for a particular zone is managed by the zone owner operator. Some zones, especially -- well, the root has certainly done this. Dot com has done it, dot SE, I believe dot UK has published policies. So there are sometimes publicly published information about how the keys are managed for a particular zone. But that really is up to each owner of the zone whether or not they choose to create the documents and, even if they create them, they may or may not make them public. It's really a zone owner choice. >> Thank you. >>JULIE HEDLUND: So we have a little less than ten minutes left and we have got a couple more questions. So we are needing to wrap it up. Did you have one in the chat? No? Okay. I saw two other questions. This gentleman and that one. So we will take those two and I think we will probably have to start wrapping up. Hang on. >> Okay. (Speaking in French) (Scribes receiving no English translation) >>ROY ARENDS: I had the first part of the question. The second part of the question, if you can please repeat that for the translator. Thank you. (Speaking in French) (Scribes receiving no English translation) >>ROY ARENDS: Okay. I think I understand the question. Let me just repeat that and see if it's correct. DNSSEC is based on authentication of DNS data. Since -- (Speaking in French) >>ROY ARENDS: Sorry. Please let me finish. Thank you. And, because you introduced that, you inherently make DNS slower. Now my response to that is, sure, you do more work at the validated part. And, yes, it's a little bit slower. But that amount of slower is negligible. Because, remember, after it validates, you can cache the data. And, if you measure the DNS, what you normally measure as well is the amount of the speediness, basically, how fast you can look up something in the cache. That doesn't need to be validated because it has been validated before. So that's one. The second part is validators at an ISP level, if it's a large constellation of systems and they have to do a lot of work, you have something similar as, for instance, as the self-terminators. That's, basically, a piece of hardware that can accelerate the validation of the DNSSEC. But, in any case, it is not that slower. The amount -- I think people, in general, talk more about the increase in traffic and the size of the packets than that they talk about slower resolving. Because I think, really, the amount of -- the amount of delay that it introduces is negligible. I hope I answered the question correctly. I hope I understood the question correctly. >>JULIE HEDLUND: Last question, and then we do have to wrap up. (Speaking in French). (Scribes not receiving English translation) >>ROY ARENDS: Okay. I will try to answer this short. There are multiple things in your question. Question one, related to your first question is: In terms of statistics, how much is the overhead? How much is the overhead DNSSEC introduces? It's negligible. There is overhead, but I think it is negligible. What I can recommend is on Wednesday morning we have the DNSSEC section, a DNS working group. Everyone here is invited and tell your friends as well. We're going to have a TLD update. It's a panel session somewhere in the middle of the whole day. During the panel session I have a few slides which will show you some statistics about the overhead and about the increase in traffic and about the uptake of DNSSEC in the U.K. And I don't have that right in my head. I did prepare for that presentation. But I recommend you come on Wednesday, and I can show you the data. Yeah, thank you. I think that wraps it up. >>JULIE HEDLUND: Yep. Thank you, everyone. And thank you very much, Roy, for mentioning the DNSSEC workshop. That is -- it starts at 8:30 on Wednesday in this room. So you just have to come back here on Wednesday morning and get a lot more information on DNSSEC. >>SIMON McCALLA: First, can I thank our fantastic translation staff over here who have done a brilliant job translating a really difficult subject. [Applause] For those who have been following. And please can I thank again the panel here for all the hard work and all the preparation. Can I say also the questions that you have asked today have been the best questions we have ever had in this session in all the times we've done it. So please come on Wednesday. We're in this room running all day, so please do come and take part again. So thank you very much, everybody. [Applause]