WEBVTT 00:00:09.210 --> 00:00:12.820 applause 00:00:12.820 --> 00:00:16.360 Karsten Nohl: Great to be back. Thank you very much, talking once again on mobile 00:00:16.360 --> 00:00:21.080 security, taking two very different angles, though, from what we talked about 00:00:21.080 --> 00:00:26.670 the last couple of years. This time we want to dive into the same topic that Tobias 00:00:26.670 --> 00:00:31.890 Engel just did, looking at insecurities that arise from the interconnect networks 00:00:31.890 --> 00:00:38.150 between different operators and we want to add another angle. And that is how YOU 00:00:38.150 --> 00:00:43.190 can start self defending yourself from the insecurities that many of your operators 00:00:43.190 --> 00:00:49.410 have left open for many years, including the new ones that Tobias and myself talk 00:00:49.410 --> 00:00:56.190 about. If you do watch this on a download, do go back and also watch Tobias's talk, 00:00:56.190 --> 00:01:00.110 it's well worth it and also covers a lot of the basics that I'm just going to skip 00:01:00.110 --> 00:01:06.320 over now for the sake of time. Great talk, by the way. Thank you Tobias. So aside 00:01:06.320 --> 00:01:17.460 from. applause Aside from those SS7 based attacks, we want to talk about 3G 00:01:17.460 --> 00:01:23.780 insecurities, not too many of them, but severe as ever, as well as in the last 00:01:23.780 --> 00:01:30.210 chapter. Then a few tips, as well as a new tool to help you start self defending 00:01:30.210 --> 00:01:36.320 against these mobile attacks. Now, just briefly, then, what is the SS7 Network 00:01:36.320 --> 00:01:40.920 Tobias has already covered the basics. So just a quick definition from me. It's this 00:01:40.920 --> 00:01:45.890 network that different mobile operators are connected to, to exchange data among 00:01:45.890 --> 00:01:51.530 each other. For instance, text messages are sent over this network. So without SS7, 00:01:51.530 --> 00:01:57.920 you couldn't be using this ancient chatting technology SMS. Thank you SS7. But also 00:01:57.920 --> 00:02:05.280 more security relevant information is exchanged over SS7. For instance, if you're 00:02:05.280 --> 00:02:10.530 using your phone in another country, as many of you currently do, you still want 00:02:10.530 --> 00:02:15.510 this visiting network to be able to use encryption with your phone, but how is that 00:02:15.510 --> 00:02:20.030 network going to know the right encryption key? So this visiting network, the German 00:02:20.030 --> 00:02:24.190 network has to ask your home network for the correct encryption key and that goes 00:02:24.190 --> 00:02:29.700 over SS7. And you can already see if there's cryptographic information being 00:02:29.700 --> 00:02:33.560 exchanged, if the wrong people ask and still receive an answer, insecurities 00:02:33.560 --> 00:02:39.950 arise. More interesting from a security perspective, though, are messages that are 00:02:39.950 --> 00:02:46.099 exchanged within one network over SS7. So SS7 is often misunderstood as this 00:02:46.099 --> 00:02:50.840 technology that's used for worldwide exchange of information. The same network, 00:02:50.840 --> 00:02:54.640 though, is used inside an operator. So there's no need for interconnect. There's 00:02:54.640 --> 00:03:01.290 already SS7 flows going on between those different mobile switching centers, MSC. 00:03:01.290 --> 00:03:07.530 And each mobile switching center covers one area, let's say a city. So imagine a 00:03:07.530 --> 00:03:13.340 situation where you are. You're in a call and you're traversing from one area to 00:03:13.340 --> 00:03:17.110 another. You're crossing, let's say, your state boundary. So there's new MSC, 00:03:17.110 --> 00:03:21.560 doesn't know how to handle your call. It needs the decryption key for the already 00:03:21.560 --> 00:03:28.610 ongoing conversation. So there's another SS7 message that allows you to query for 00:03:28.610 --> 00:03:33.650 the key of a transaction that's currently going on. OK? And again, you can already 00:03:33.650 --> 00:03:38.739 see how if the wrong people send this type of message and they receive an answer, 00:03:38.739 --> 00:03:46.730 insecurities arise. The insecurity that that has most been talked about in recent 00:03:46.730 --> 00:03:52.670 years, again, up until Tobias's talk, was tracking. And tracking was often understood 00:03:52.670 --> 00:03:57.720 as: There's this evil message, the any time interrogation and The Washington Post 00:03:57.720 --> 00:04:01.621 focused a lot an article on just one message. And it's a it's really evil. It 00:04:01.621 --> 00:04:06.451 should not been I have been ever standardized. And whenever it's used, it's 00:04:06.451 --> 00:04:12.101 for evil purposes. There's no usefulness in this message. And Tobias 00:04:12.101 --> 00:04:16.209 quoted a number that I think The Washington Post found in a lot of 00:04:16.209 --> 00:04:21.709 marketing material, 70 percent of mobile networks respond to this message. Now, 00:04:21.709 --> 00:04:26.000 this is information from earlier this year. A lot of networks, very good news, have 00:04:26.000 --> 00:04:32.770 moved to to stop responding to anytime interrogation message. This evil spying 00:04:32.770 --> 00:04:37.509 message is not being responded to by, for instance, all German networks. You can't 00:04:37.509 --> 00:04:44.460 use this message in Germany anymore. However, this is a very retroactive 00:04:44.460 --> 00:04:51.729 approach to securing SS7 because there's a number of other messages that, consider them 00:04:51.729 --> 00:04:57.169 Gadgets, get you to the same place, take a phone number and take you all the way to 00:04:57.169 --> 00:05:03.439 somebody's location. And here's just a snapshot of of which messages you can use 00:05:03.439 --> 00:05:08.960 and Tobias went into a greater level of detail in how these different messages 00:05:08.960 --> 00:05:13.689 come together. So if anybody thinks that just barring anytime integration, you 00:05:13.689 --> 00:05:20.642 solved the tracking problem, they are wrong. But at the same time, it's not that SS7 is 00:05:20.642 --> 00:05:26.900 not secureable. It's just a much larger challenge that people consider currently 00:05:26.900 --> 00:05:33.869 to be. So you see how stringing together some of these messages get you to 00:05:33.869 --> 00:05:39.039 intermediate values that also shouldn't be public and then all the way to a cell ID. 00:05:39.039 --> 00:05:42.849 And up until all these messages or at least every path that takes you from left 00:05:42.849 --> 00:05:49.759 to right is blocked by a network, tracking to the same accuracy, to cell ID stays 00:05:49.759 --> 00:05:54.949 possible. Now, this is just one of many areas in which SS7 can become an issue. 00:05:54.949 --> 00:06:03.559 Here is 4 more, it's an intercept risk. If people can read your SMS text or listen 00:06:03.559 --> 00:06:08.169 to your calls, it's a denial of service risk. If people cut you off from 00:06:08.169 --> 00:06:13.490 phone connectivity for anywhere from an hour until the next location update or 00:06:13.490 --> 00:06:19.319 until your next reboot your phone, so you can really cut people off badly from it, 00:06:19.319 --> 00:06:24.559 from the phone network. This area of fraud that I don't think many people want to 00:06:24.559 --> 00:06:29.249 talk about publicly, certainly I don't. But there's many fraud risks in SS7 00:06:29.249 --> 00:06:34.089 in which you can easily put charges on somebody else's bill, or more 00:06:34.089 --> 00:06:39.899 interestingly, you can remove limits on your own prepaid cards, basically run up 00:06:39.899 --> 00:06:46.240 infinite charges on prepaid cards and, you know, running up a lot of bills to a two 00:06:46.240 --> 00:06:50.960 to premium numbers, for instance. And then there's the risk of spamming, which from 00:06:50.960 --> 00:06:55.930 what I hear is already happening, SS7 based spam attacks. Now, for the sake of 00:06:55.930 --> 00:07:01.560 this talk, I want to focus on intercept, which I consider aside from tracking the 00:07:01.560 --> 00:07:06.099 most intrusive and the most relevant for us, just as a risk, they're more relevant 00:07:06.099 --> 00:07:09.649 for the network operators. And if they don't solve them, well, so be it, as long 00:07:09.649 --> 00:07:14.469 as they foot the bill for it. So intercept. And I want to go into three 00:07:14.469 --> 00:07:21.250 possible scenarios in which SS7 assisted intercept can happen. The first abuses 00:07:21.250 --> 00:07:24.719 the exact message, as we looked at in the introduction, these messages where 00:07:24.719 --> 00:07:29.889 different parts of networks ask each other for encryption information and it's a 00:07:29.889 --> 00:07:35.860 pretty straightforward attack. You record the airwaves. Around somebody in 00:07:35.860 --> 00:07:41.129 somebody's vicinity and you record somebody's encrypted transaction as part of 00:07:41.129 --> 00:07:47.050 that, right? So and 3G transaction, for instance, are pretty well secured, but 00:07:47.050 --> 00:07:53.080 they're not very hard to record. In fact, 3G is a little bit easier than 2G because 00:07:53.080 --> 00:07:58.039 it doesn't jump around all these frequencies. So you record, let's say, 3G 00:07:58.039 --> 00:08:02.949 data and you have a bunch of transactions. And all of them encrypted. And you can use 00:08:02.949 --> 00:08:09.939 this message over SS7 to decrypt them. It's called Send ID. And as a as I said on 00:08:09.939 --> 00:08:16.129 one of the earlier slides, it's supposed to be used when you're moving from one MFC 00:08:16.129 --> 00:08:20.810 into another MSC, but still within your own network so that the call doesn't get 00:08:20.810 --> 00:08:27.099 disrupted. It's not supposed to be used when when somebody foreign wants to 00:08:27.099 --> 00:08:31.779 query your phone, if they need a new encryption key, a new call needs to start 00:08:31.779 --> 00:08:36.270 anyway. There's no way to hand over a call from one operator to another operator 00:08:36.270 --> 00:08:43.209 without disruption. So this message is used only for internal purposes. However, 00:08:43.209 --> 00:08:47.780 out of the four German operator earlier this month, all four responded to this 00:08:47.780 --> 00:08:52.100 request coming from another country, another country that doesn't even border 00:08:52.100 --> 00:08:57.170 Germany. So there's no way to even conceptually think a call would be handed 00:08:57.170 --> 00:09:03.950 over. So four out of four. And that's not an anomaly. Most networks require an 00:09:03.950 --> 00:09:08.940 international response to an outside number when asked for the current 00:09:08.940 --> 00:09:14.030 decryption key. I'll show you a quick demo on this at the end of this chapter. 00:09:14.030 --> 00:09:17.650 But I first finish the enumeration of all the different possibilities in which 00:09:17.650 --> 00:09:24.920 3G calls can be intercepted. The second one, the good old IMSI catchers, which we 00:09:24.920 --> 00:09:31.540 also wouldn't work on 3G. And I guess for the most part they don't unless SS7 00:09:31.540 --> 00:09:36.010 comes to the help. So why don't they work without SS7? An IMSI catcher 00:09:36.010 --> 00:09:42.070 pretends to be a base station. And if it's 2G technology, the phone has no way 00:09:42.070 --> 00:09:47.720 of knowing the difference between the real base station and a fake base station. But 00:09:47.720 --> 00:09:53.180 then 3G, the 3G standard introduced what I call mutual authentication. So this time 00:09:53.180 --> 00:09:57.630 the base station has to prove to a phone that in fact it's legitimate and unless it 00:09:57.630 --> 00:10:03.530 does that, the phone won't connect. Now, this only solves part of the IMSI catcher 00:10:03.530 --> 00:10:08.530 problem. Just taken by the name even the catching is still possible, IMSI catching 00:10:08.530 --> 00:10:14.660 in the sense of creating a list of all the IMSIs in a location. Because there's 00:10:14.660 --> 00:10:19.150 certain chicken and egg problem. If you want me as a base station to 00:10:19.150 --> 00:10:23.430 authenticate to you, you first have to tell me who you are. There's no such thing 00:10:23.430 --> 00:10:28.370 as SSL or any type of public key on the mobile network. It's all symmetric key. So 00:10:28.370 --> 00:10:32.900 you first have to tell me which key to use and by that I know who you are. So IMSI 00:10:32.900 --> 00:10:36.811 catching is always possible. And that's why if you Google for 3G IMSI catcher, those 00:10:36.811 --> 00:10:43.240 things exist. But they aren't capable of recording phone calls or SMS because those 00:10:43.240 --> 00:10:49.080 then required a mutual authentication. They aren't capable of doing so unless they ask 00:10:49.080 --> 00:10:55.960 over SS7 for an authentication key. So IMSI catchers are back in the 3G world 00:10:55.960 --> 00:11:05.328 big time, unless we solve these SS7 problems, right? The third possibility of 00:11:05.328 --> 00:11:10.880 of intercept - this is probably the scariest because it can happen completely 00:11:10.880 --> 00:11:15.470 remotely - Boaster once enumerated so far, you have to be somewhere in the vicinity 00:11:15.470 --> 00:11:19.540 in the vicinity of somewhere. So the third possibility, I want to call the rerouting 00:11:19.540 --> 00:11:24.640 attacks and they work in both directions. Rerouting is the idea. And to be as 00:11:24.640 --> 00:11:31.270 touched on this, of taking… of taking somebodies phone calls and changing 00:11:31.270 --> 00:11:36.799 the destination number so that, in fact, you call somebody else unbeknownst to you, 00:11:36.799 --> 00:11:42.590 of course, as the victim. And this will expose for incoming calls and outgoing 00:11:42.590 --> 00:11:46.600 calls, but using very different methods. So it just kind of accidentally works in 00:11:46.600 --> 00:11:52.970 both directions. And this part, I just briefly want to demonstrate to BSN that 00:11:52.970 --> 00:11:56.870 coordinated on most of this. But this part, I guess we kind of misunderstood 00:11:56.870 --> 00:12:01.870 each other as we both showed us. I'll keep this very brief. And the point I want 00:12:01.870 --> 00:12:07.998 to get across is that, one, a single SS7 message is already a big intercept 00:12:07.998 --> 00:12:15.660 problem. Let's see. Connected here. Um, so I'll try not to make the same mistake as 00:12:15.660 --> 00:12:26.600 Tobias and try to cut off part of my number here. So 31C3 demo phone. 00:12:26.600 --> 00:12:32.713 So I'm calling a a phone that in fact, accidentally we left in. So … fuck NOTE Paragraph 00:12:32.713 --> 00:12:36.190 Laughter and applause Ring-back tone starts 00:12:36.190 --> 00:12:40.491 So I am calling this number and I don't know if you can hear it, but it's ringing. 00:12:40.491 --> 00:12:43.813 And we did leave his phone back in Berlin accidentally. But for the sake of this 00:12:43.813 --> 00:12:48.100 demo, that makes no difference. So it's a it's a phone somewhere in Berlin. Nobody 00:12:48.100 --> 00:12:50.912 answers to. Here is another phone. 00:12:50.912 --> 00:12:52.002 Ring-back tone stops 00:12:52.002 --> 00:12:54.329 So if I if I register what they call a 00:12:54.329 --> 00:13:01.220 supplementary service to this number. And that's just fancy language for, for, for 00:13:01.220 --> 00:13:09.392 call forwarding, if I call this exact same number again. 00:13:13.758 --> 00:13:16.659 Ring-back tone starts 00:13:16.659 --> 00:13:18.877 Phone ringing also starts 00:13:18.877 --> 00:13:21.140 This phone is ringing. 00:13:21.140 --> 00:13:23.930 Applause 00:13:23.930 --> 00:13:25.800 Both ring-back and ring-tone stop 00:13:25.800 --> 00:13:28.059 Still applause 00:13:28.059 --> 00:13:33.120 Now, of course, to make this real intercept, I wouldn't forward it to a 00:13:33.120 --> 00:13:37.740 phone, I would forward it to a computer that then is smart enough to very quickly 00:13:37.740 --> 00:13:43.960 erase the call forwarding and call the original number and then connect it to so 00:13:43.960 --> 00:13:48.260 that the phone, the phone call actually goes to where it was supposed to go. Just 00:13:48.260 --> 00:13:53.451 I'm sitting in the middle and I'm receiving a copy of it. OK, so that's the 00:13:53.451 --> 00:13:57.710 idea in this direction, in the other direction, the exact same thing works as 00:13:57.710 --> 00:14:03.875 well. And Tobias already told you how these services that say, let me rewrite 00:14:03.875 --> 00:14:07.510 your phone number for you because you don't know how to dial a phone number when 00:14:07.510 --> 00:14:12.279 you're on vacation. Right. Those services can be set by anybody, at least on a lot 00:14:12.279 --> 00:14:16.880 of networks. And you can see how the exact same thing works there so that every time 00:14:16.880 --> 00:14:21.430 you dial a number that just move their own number in place of that number and then 00:14:21.430 --> 00:14:26.912 connect those two calls. So, as I said, I consider those to the scariest type of 00:14:26.912 --> 00:14:30.680 attacks because they were completely remotely you don't have to be in the radio 00:14:30.680 --> 00:14:35.140 vicinity of anybody. And surprisingly, this still works against a bunch of 00:14:35.140 --> 00:14:41.690 networks, even against those networks that move to solve some of the earlier issues. 00:14:41.690 --> 00:14:49.285 So networks [are] still very retroactive. So what do what do those mobile networks 00:14:49.285 --> 00:14:54.920 now have to do to to solve those issues? Well, as always, of course, the answer: 00:14:54.920 --> 00:14:59.921 It depends. It depends in this case on the tech type. Some of the techs can simply be 00:14:59.921 --> 00:15:05.710 blocked. Like the AnytimeInterrogation, that earlier this year they said 70% of 00:15:05.710 --> 00:15:10.170 the networks are vulnerable. Now in Germany it's zero. So something happened 00:15:10.170 --> 00:15:16.440 there. And the same is true for the for the first type of attack that I've shown. 00:15:16.440 --> 00:15:20.550 The passive intercept I said when we tested earlier this month for other four 00:15:20.550 --> 00:15:27.100 networks are vulnerable. Now it's down to two. So within two weeks, two networks put 00:15:27.100 --> 00:15:33.970 in a firewall rule that says this message has no purpose. Traversing our outside 00:15:33.970 --> 00:15:39.940 network boundary, just block it. The typical firewall is the same isn't 00:15:39.940 --> 00:15:45.100 possible for these other two types of attacks because those messages are 00:15:45.100 --> 00:15:50.550 actually useful. They do something, at least in certain circumstances. If you 00:15:50.550 --> 00:15:55.210 block the second type of query here to send authentication info, you couldn't be 00:15:55.210 --> 00:15:58.930 roaming in another country anymore. If you blocked a third one, you couldn't be 00:15:58.930 --> 00:16:04.400 changing your your voice mail forwarding from another country anymore. So these are 00:16:04.400 --> 00:16:10.390 needed. Still we couldn't, we can't accept that just anybody who asks over SS7 ... 00:16:10.390 --> 00:16:11.990 Phone ringing Nohl sighs 00:16:11.990 --> 00:16:15.658 You guys! Laughter 00:16:15.658 --> 00:16:23.750 Switched this off. We can't accept that just anybody who asks over SS7 00:16:23.750 --> 00:16:29.370 receives an answer, at the very least we would expect networks to only answer to 00:16:29.370 --> 00:16:33.500 their friends on SS7, and that is their roaming partners. That's 00:16:33.500 --> 00:16:38.980 already a lot fewer companies and especially a lot fewer sketchy companies 00:16:38.980 --> 00:16:44.791 than everybody else on SS7. We would then want those networks to do some 00:16:44.791 --> 00:16:51.390 plausibility checking. Right. So this does phone in Berlin that just put a 00:16:51.390 --> 00:16:56.670 supplementary service on. The network operator knows the phone is in Berlin and 00:16:56.670 --> 00:17:02.760 I send us from the other end of the world. Still, they are not on it. Right. Any type 00:17:02.760 --> 00:17:08.310 of possibility checking what would clearly see that this is not possible for a phone 00:17:08.310 --> 00:17:12.760 to be in one country and for this user to want to change their voicemail setting 00:17:12.760 --> 00:17:17.809 from somewhere completely different. And then thirdly, networks need to limit the 00:17:17.809 --> 00:17:22.020 rate at which this happens. Those services that The Washington Post talked about is 00:17:22.020 --> 00:17:26.240 tracking services. These are large operations. They seem to be tracking 00:17:26.240 --> 00:17:33.620 thousands of people, constantly. This will show in logs, you don't allow some random 00:17:33.620 --> 00:17:38.300 network somewhere else in the world to constantly interrogate hundreds of your 00:17:38.300 --> 00:17:44.200 users, right? It's clearly abuse. Has any network move to put such sensible rules 00:17:44.200 --> 00:17:48.429 in? I'm not aware of it, but it's certainly the next step. And I'm not ready 00:17:48.429 --> 00:17:54.860 to give up on SS7 yet. I've heard one too many times that SS7 is an old technology 00:17:54.860 --> 00:18:01.389 built with no security in mind and we just can't fix it. The Internet also is an old 00:18:01.389 --> 00:18:06.399 technology built was not secured in mind, and we did fix it since the 90s, since 00:18:06.399 --> 00:18:10.679 when you connected to Windows 95 computer to the Internet, it got infected with the 00:18:10.679 --> 00:18:16.580 virus right away. We have moved to put in firewalls. We're not exposing our printer 00:18:16.580 --> 00:18:21.190 daemon and now file-sharing daemon on the entire Internet anymore for four billion 00:18:21.190 --> 00:18:25.680 people to connect to and the same as possible on SS7. Which is, we we're still 00:18:25.680 --> 00:18:34.508 in the nineties. Thank you. Applause 00:18:34.508 --> 00:18:38.484 Having said that though, let me show you what what happens if we don't do that, 00:18:38.484 --> 00:18:46.972 the fun part. So. We argued whether or not we wanted to show this as a live demo. 00:18:46.972 --> 00:18:50.096 You'll understand why we don't show it as a live demo. There is just too much stuff 00:18:50.096 --> 00:18:54.470 that could go wrong. But here's the setup. We start with just a phone number 00:18:54.470 --> 00:19:00.389 and we want to string together a couple of SS7 gadgets while also having this radio 00:19:00.389 --> 00:19:05.105 handy that can capture 3G information to capture yet more information that's not 00:19:05.105 --> 00:19:10.870 available over SS7. Right. So we start with a phone number and we send what's 00:19:10.870 --> 00:19:18.195 called an SRI-for-SM message, which gives us, if the network is configured answer, 00:19:18.195 --> 00:19:26.441 the IMSI and the MSI that the subscriber currently is connected for. Those two are 00:19:26.441 --> 00:19:31.001 used as parameters into another call. Called the PSI message, provide 00:19:31.001 --> 00:19:37.191 subscriber info. And then that call then gives us the Cell ID. This is just how 00:19:37.191 --> 00:19:41.440 you get more and more information with different gadgets. Now the Cell ID tells 00:19:41.440 --> 00:19:45.840 us where somebody is physically. So imagine we now move our radio to that 00:19:45.840 --> 00:19:54.309 location and we again send a PSI. We record the PSI. We set radio, not the PSI, what 00:19:54.309 --> 00:19:59.779 happens over the airways when we send the PSI and the phone gets paged. So when we 00:19:59.779 --> 00:20:05.889 send the PSI over SS7, the phone receives some information. Right. This radio plus a 00:20:05.889 --> 00:20:11.070 little bit GNU radio scripting gives us that information: Who has been paged 00:20:11.070 --> 00:20:18.749 during that short window of time that we that we recorded? Now when we record 00:20:18.749 --> 00:20:22.929 something on UMTS, we always record for different cells – they share frequencies. 00:20:22.929 --> 00:20:27.419 But you see that the one cell with the Cell ID came back over SS7 is included 00:20:27.419 --> 00:20:33.012 in our set. So we filter the data for that cell and we look for which IMSIs are 00:20:33.012 --> 00:20:36.739 included. And luckily for us, only one IMSI got paged within those few 00:20:36.739 --> 00:20:43.490 seconds on that cell. It's the same. Same. This is now the TMSI that belongs to 00:20:43.490 --> 00:20:48.600 this phone. This is information we can't get over SS7. But what you can do over SS7 00:20:48.600 --> 00:20:54.710 with the TMSI is request a key, so it gets complicated. But so we have the decryption 00:20:54.710 --> 00:21:00.250 key now and the next time this phone receives something, unless it changes the 00:21:00.250 --> 00:21:04.500 key, in which case we can ask again for a new key. Next time this phone receives 00:21:04.500 --> 00:21:07.279 something. And what you don't see in the video is, somebody is now sending a text 00:21:07.279 --> 00:21:12.129 message to the phone. We can also record that right. Again, same radio, the one 00:21:12.129 --> 00:21:17.990 shown in the picture, now the phone that received a text message. And there's a few 00:21:17.990 --> 00:21:26.980 more steps. So the phone received a text message and we also, again, recorded the 00:21:26.980 --> 00:21:38.629 airwaves. We again run it through some GNU radio script. Now, was was UMTS 00:21:38.629 --> 00:21:42.529 everything? It is kind of complicated, so there's a different connection, of 00:21:42.529 --> 00:21:45.779 course, happening all at the same time, and then they'll get allocated to 00:21:45.779 --> 00:21:49.999 different channels. So now, in order to to decode this text message, we're going to 00:21:49.999 --> 00:21:55.950 find out which channel is used. So this command gives us the list of which which 00:21:55.950 --> 00:22:00.909 channels have been allocated. And we got to find a TMSI from earlier in one of 00:22:00.909 --> 00:22:06.040 these channel allocations. And Wireshark is a great help in this. We didn't have to 00:22:06.040 --> 00:22:11.050 do anything with Wireshark. I just knows all that 3G stuff right out of the box. So 00:22:11.050 --> 00:22:14.970 luckily, the first of these five connecting requests is the right one and 00:22:14.970 --> 00:22:19.379 scroll all the way down, there's then the parameters that say which channel this 00:22:19.379 --> 00:22:23.919 transaction happened on. So those two numbers, 15 and 48 is the channel. So we, 00:22:23.919 --> 00:22:31.324 we need to cell frequency, but we need those those two two numbers, that, that 00:22:31.324 --> 00:22:36.749 are the channel and the key, you know, this is only 64 bit. I'll discuss that 00:22:36.749 --> 00:22:46.675 a little later. And that's all we need to decrypt an SMS. And there it is. 00:22:46.675 --> 00:22:55.382 Applause Thank you. 00:22:57.359 --> 00:23:03.540 This still works today, but only against two out of the four German networks. Some 00:23:03.540 --> 00:23:10.351 of them move to to to stop some of these messages, of course, most importantly, 00:23:10.351 --> 00:23:14.940 this SI message that gives you the decryption key. But even if you block this 00:23:14.940 --> 00:23:22.539 message, just acquiring somebody's location can already be intrusive enough. 00:23:22.539 --> 00:23:27.389 All right. Moving on to 3G security or rather extending on 3G security since this 00:23:27.389 --> 00:23:34.919 already touched through 3G in a big way. You remember the good old days where where 00:23:34.919 --> 00:23:40.489 you could just intercept all phone calls was the Osmocon phone. Thank you, by the 00:23:40.489 --> 00:23:45.059 way, for that open source project that helped us so much over the years. And you 00:23:45.059 --> 00:23:52.849 combine that with the kraken software to decrypt the phone call. So with 20 year 00:23:52.849 --> 00:23:57.919 old vers of phone and the server you can listen to anybody's GSM calls as long as 00:23:57.919 --> 00:24:03.940 they're using the A5/1 cipher. Some networks recently moved into A5/3. 00:24:03.940 --> 00:24:10.720 So it doesn't work this way anymore. Now, how does this now compare to 3G security? 00:24:10.720 --> 00:24:16.039 As I've just shown, basically the same attacks are possible. Instead of the 00:24:16.039 --> 00:24:21.419 Osmocom phone, we use a programable radio, some more software, but again, very 00:24:21.419 --> 00:24:26.509 affordable 400 euros or something. And you combine that using 00:24:26.509 --> 00:24:34.409 instead of kraken SS7 queries. So unless we fix SS7, 3G is no more secure than 2G 00:24:34.409 --> 00:24:41.460 and neither is A5/3, the recent upgrade of GSM because those keys are 00:24:41.460 --> 00:24:50.500 again exposed over SS7. Now, some networks, you don't even need that second 00:24:50.500 --> 00:24:57.559 part, so they have bigger things to worry about and then SS7 attacks and our data 00:24:57.559 --> 00:25:01.919 set isn't all that large. Some of you provided measurements through through a 00:25:01.919 --> 00:25:07.260 software release last year. So thank you very much for that. And we have captures 00:25:07.260 --> 00:25:14.619 from maybe 20, 25 countries out of those five having to use no 3G encryption at 00:25:14.619 --> 00:25:21.200 all. Well, four countries. Five network operators. Right. Which I find shocking. 00:25:21.200 --> 00:25:26.249 Some of these even have encryption turned on on their GSM network and then forgot to 00:25:26.249 --> 00:25:31.216 turn it on or deliberately left it out because it's harder to intercept on the 3G 00:25:31.216 --> 00:25:38.330 variant. Right. So those networks, as I said, have much more, much more worrisome 00:25:38.330 --> 00:25:45.350 issues than SS7 attacks. And they really need to be called out. And we do that with 00:25:45.350 --> 00:25:49.659 an extension of a website that we've been maintaining for a couple of years, gsmmap, 00:25:49.659 --> 00:25:55.860 big update of gsmmap launched today with all the 3G measurements, we, we 00:25:55.860 --> 00:26:01.590 collected and you collected over the last couple of years. Now, some of you may have 00:26:01.590 --> 00:26:07.951 used gsmmap before. The idea as to to rank operators in the three categories. How 00:26:07.951 --> 00:26:13.509 hard is it to intercept phone calls and SMS? Is it easy to impersonate a person 00:26:13.509 --> 00:26:17.950 and then put charges on a bill, for instance, or receive the calls? How hard 00:26:17.950 --> 00:26:22.760 is it to track them? And as you see, over the last years, networks have improved 00:26:22.760 --> 00:26:31.220 their security, at least some, as always. God. And as you also see, these are the 2G 00:26:31.220 --> 00:26:39.049 networks, even the best secure 2G network. And in Germany anyway, in our opinion, is 00:26:39.049 --> 00:26:44.450 less secure than the worst secured 3G networks. These are for 3G networks, still 00:26:44.450 --> 00:26:50.399 we want networks to implement all security features. And as you saw before, some 00:26:50.399 --> 00:26:57.399 other countries don't have that luxury of all 3G secure networks reasonably secure. 00:26:57.399 --> 00:27:01.909 Not the first version of our metric is very crude and we want to improve upon 00:27:01.909 --> 00:27:06.210 this over time. But currently how we calculate the score is we'll give ninety 00:27:06.210 --> 00:27:10.779 percent of the points to anybody who switches on encryption. That's the main 00:27:10.779 --> 00:27:16.330 security feature and the remaining 10 percent you earn by changing the TMSI 00:27:16.330 --> 00:27:22.149 quickly. TMSI is what we needed for these SS7 attacks to work well. So if you keep 00:27:22.149 --> 00:27:28.440 changing it, it really confuses the that the person trying to to haunt you also 00:27:28.440 --> 00:27:32.559 this makes other types of attacks more difficult, will factor in a couple of more 00:27:32.559 --> 00:27:38.989 values as we collect more data. But this is it for now. So, yeah, big update on 00:27:38.989 --> 00:27:43.880 gsmmap. If you haven't checked it out, check out your country on gsmmap, read the 00:27:43.880 --> 00:27:52.149 country report. So does a six page or so report, auto generated, that explains what 00:27:52.149 --> 00:27:56.759 types of measurements we included into into these graphs and why we think they 00:27:56.759 --> 00:28:01.529 they constitute certain risks. Maybe forward it to to your network and say if 00:28:01.529 --> 00:28:08.870 you're not improving, I'm going to change, switch to another network. Now, not 00:28:08.870 --> 00:28:14.210 everything is on, on gsmmap yet because we don't have enough data. And there's one 00:28:14.210 --> 00:28:19.080 problem in particular that I want to start warning about, because I really think 00:28:19.080 --> 00:28:24.399 we're running into an issue here. And that is the lengths of encryption key you saw 00:28:24.399 --> 00:28:29.759 in the in the capture, in the video data that I showed that the key that came back 00:28:29.759 --> 00:28:37.419 over SS7 was actually only 64bit from this particular network. And the SIM card that 00:28:37.419 --> 00:28:41.440 was there was used in this attack, was bought that very same week. So we recorded 00:28:41.440 --> 00:28:46.039 this video last week. So it's the the most recent SIM card you can buy from this 00:28:46.039 --> 00:28:51.340 network. And still it only uses 64 bit. And that, in my view, is incompatible with 00:28:51.340 --> 00:28:57.710 what we have learned from from recent Snowden documents that the NSA in 2011, 00:28:57.710 --> 00:29:06.149 2012 funded a project to break A5/3. This is a 64 bit cipher. And we had 00:29:06.149 --> 00:29:09.919 estimated at this very conference a year ago that you'd need about a million 00:29:09.919 --> 00:29:14.759 dollars to break A5/3. Now, they did it a little bit earlier. So Moore's 00:29:14.759 --> 00:29:19.300 Law, everything's more expensive and probably to have overhead, too. But they 00:29:19.300 --> 00:29:25.000 spend apparently four billion pounds. I don't know why pound, not dollars, but it 00:29:25.000 --> 00:29:31.200 may have been some GCHQ Corporation. So for four million pound a couple of years 00:29:31.200 --> 00:29:36.791 ago, you could already break 64 bit crypto and 64 bit is more prevalent in mobile 00:29:36.791 --> 00:29:44.499 networks than you would have thought when they upgraded the GSM networks to A5/3. 00:29:44.499 --> 00:29:49.342 They didn't actually upgraded it to UMTS security, as everybody claimed they did. 00:29:49.342 --> 00:29:57.771 They upgraded it to the cipher used in UMTS with a key half the size. When 00:29:57.771 --> 00:30:02.958 writing the A5/3 standards though, the people were smart enough to also put in 00:30:02.958 --> 00:30:10.669 the real UMTS cipher with full key size, they called it A5/4 and it has never 00:30:10.669 --> 00:30:15.029 been seen anywhere since. It's written in the standard. It was released the same day 00:30:15.029 --> 00:30:20.960 that A5/3 was released. Nobody has ever moved to implement that. So GSM for the 00:30:20.960 --> 00:30:26.049 time being is and will be vulnerable to anybody. It was a one million dollar 00:30:26.049 --> 00:30:30.911 machine in the basement. Certainly NSA, but more and more people as we move 00:30:30.911 --> 00:30:34.570 forward. And what costs a million dollars today, thanks to Moore's Law in a couple 00:30:34.570 --> 00:30:40.869 of years, anybody can break it on a computers like we today. Break the A5/1. 00:30:40.869 --> 00:30:45.649 If your network uses certain older SIM cards, differentiation years between a 00:30:45.649 --> 00:30:52.529 SIM card and a USIM as a UMTS SIM card. If your network only uses SIM cards, then 00:30:52.529 --> 00:30:59.590 even your 3G transactions are 64 bit encrypted. So there is no way to generate 00:30:59.590 --> 00:31:02.960 more entropy. You could query for two keys, I guess, but they weren't smart 00:31:02.960 --> 00:31:10.730 enough to do that. So 64 bit encryption for UMTS and that's just not good enough. 00:31:10.730 --> 00:31:15.309 And as I said, the network that we did the demo with we were surprised to see a 00:31:15.309 --> 00:31:20.700 64 bit key. We went back in our database of SIM cards. We found a lot of SIM cards 00:31:20.700 --> 00:31:25.027 that have this problem. We want to add this to gsmmap, but we don't want to be 00:31:25.027 --> 00:31:29.214 unfair just because we see one very old SIM card in the network. We don't want to give 00:31:29.214 --> 00:31:32.987 them a low score versus somebody else, where we only see a new card. So we need 00:31:32.987 --> 00:31:38.596 lots and lots of data. Help us collect those data and we'll make it public. 00:31:38.596 --> 00:31:44.345 Now, that's one reason why we stay on this ball and progress the research. The other 00:31:44.345 --> 00:31:49.405 main reason, and this is really what keeps us awake at night is this question of 00:31:49.405 --> 00:31:57.120 how can we get out of the mess. We've been producing more and more problems. I should 00:31:57.120 --> 00:32:02.679 not say produce, we make you aware of more and more problems over the years and we 00:32:02.679 --> 00:32:06.570 always criticize that at least many networks do not respond to those. So we 00:32:06.570 --> 00:32:11.860 have to stockpile ever growing stockpile of mobile security issues and nobody seems 00:32:11.860 --> 00:32:15.889 to be addressing. And all we do is wait for our networks to do something 00:32:15.889 --> 00:32:20.630 eventually. Now waiting's over for me, at least I'm impatient. I want to do 00:32:20.630 --> 00:32:25.789 something now and I want to address all these issues all at once. Those issues 00:32:25.789 --> 00:32:31.169 that we talked about for several years now, including the SIM card attacks from 00:32:31.169 --> 00:32:39.739 last year, silent SMS based tracking the SMS, the SS7 abuse discussed today, 00:32:39.739 --> 00:32:46.340 IMSI Catcher Vulnerabilities and insufficiently configured networks, 2G as 00:32:46.340 --> 00:32:53.250 well as 3G. All of these problems have one thing in common. Your phone technically 00:32:53.250 --> 00:32:58.269 knows that these attacks are happening and your phone technically knows that a 00:32:58.269 --> 00:33:03.999 network is configured insecurely. But unfortunately it's buried very deep inside 00:33:03.999 --> 00:33:07.869 the phone. It's buried inside the baseband. So as much as you can program 00:33:07.869 --> 00:33:12.259 Android, you don't get access to that information. At least so we saw it and 00:33:12.259 --> 00:33:16.769 then we set out and just took the better part of this year. We wanted to dig the 00:33:16.769 --> 00:33:21.019 information out from these phones. It's somewhere in there. There must be some way 00:33:21.019 --> 00:33:27.321 to hack it out of it. And we found debug possibilities for Qualcomm chipsets, just 00:33:27.321 --> 00:33:31.309 one vendor, but extremely popular. Right now. There seem to be in every LTE phone 00:33:31.309 --> 00:33:36.809 and in a bunch of other phones. And we found, we found ways of producing exactly 00:33:36.809 --> 00:33:42.539 all the data on the right hand side to make it accessible through an Android 00:33:42.539 --> 00:33:48.060 application. And we also wrote an application for you. So: Release today. 00:33:48.060 --> 00:33:57.695 Applause 00:33:57.695 --> 00:34:05.139 Thank you, released today, SnoopSnitch under GPL. A tool that collects all the 00:34:05.139 --> 00:34:09.860 baseband information mostly to keep it on the phone and run some analysis on it, 00:34:09.860 --> 00:34:15.320 warn you about, as I said, SIM card attacks, but also those SS7 attacks that 00:34:15.320 --> 00:34:19.750 Tobias and I talked about today. How do you take those those attacks? Well, by the 00:34:19.750 --> 00:34:24.820 pagings, I showed you in the video that every time we send certain queries to 00:34:24.820 --> 00:34:30.169 the phone, to, over SS7, that the phone actually also receives information useful 00:34:30.169 --> 00:34:35.120 for the attacker. Also useful for the defender. If those empty pagings, we call 00:34:35.120 --> 00:34:38.990 them, are received by the phone, strong evidence that somebody is messing with you 00:34:38.990 --> 00:34:46.890 over SS7. Right. So it collects all that information and it produces warnings. You 00:34:46.890 --> 00:34:52.624 can also upload information issues, so you choose. It's optional of course, it runs, 00:34:52.624 --> 00:34:57.310 as I said, on a bunch of Android phones that are currently popular. It requires a 00:34:57.310 --> 00:35:01.603 somewhat recent Android version we haven't tested was Android 5 yet, but I don't 00:35:01.603 --> 00:35:05.170 see why it wouldn't work, though. We just have to put the time and your phone needs 00:35:05.170 --> 00:35:11.240 to be routed. So we have access to a certain interface that otherwise is not 00:35:11.240 --> 00:35:16.270 accessible. And it needs of course, a Qualcomm chipset, which, as you see by 00:35:16.270 --> 00:35:21.650 this list, is in most current flagship phones. It's on Google Play right now. So 00:35:21.650 --> 00:35:29.080 download it if you're interested. Now, how does this tool work? One example only, of 00:35:29.080 --> 00:35:34.500 course, right, read the source code if you if you want to know the rest. If you, for 00:35:34.500 --> 00:35:38.750 instance, IMSI catcher detection. There have been a bunch of tools so far to do 00:35:38.750 --> 00:35:43.980 IMSI catcher detection. The one we released a couple of years ago was called CatcherCatcher, 00:35:43.980 --> 00:35:49.740 but it had two limitations. One practical, one more bound to experience. 00:35:49.740 --> 00:35:54.790 The practical limitation was that it ran on Osmocom phones and Osmocom phones can't 00:35:54.790 --> 00:35:59.120 do most phone functionality. So always your second phone? And it had to be 00:35:59.120 --> 00:36:03.350 connected to a computer. So very unlikely that you carried this around all the time. 00:36:03.350 --> 00:36:07.411 And we wanted to move it onto a real phone that you can use onto your phone. Right? I 00:36:07.411 --> 00:36:11.690 think we succeeded in that. The second limitation was that we really didn't know 00:36:11.690 --> 00:36:16.440 how IMSI catchers behaved or we also didn't know how real networks behaved. And 00:36:16.440 --> 00:36:20.640 thanks to all the data on gsmmap, we think we have a much better understanding now of 00:36:20.640 --> 00:36:24.880 all the weird corner cases, how real networks behave and created a much better 00:36:24.880 --> 00:36:32.890 ruleset for for an Android based catcher catcher tool now. And the rules go in two 00:36:32.890 --> 00:36:37.111 categories. One is the configuration of the of these different cells. For 00:36:37.111 --> 00:36:41.760 instance, the lack of encryption when, you know, from the gsmmap database that this 00:36:41.760 --> 00:36:46.473 network does usually support encryption, that's a big red flag. Also certain other 00:36:46.473 --> 00:36:51.180 configurations. So that's a configuration of the network, the adjusted behavior and 00:36:51.180 --> 00:36:53.800 the IMSI catcher wants to get information out from you at the very 00:36:53.800 --> 00:36:58.290 least, the IMSI, of course, it's in the name. Right. So that suspicious behavior 00:36:58.290 --> 00:37:04.955 now, none of these things taken by themselves did allow you to detect an 00:37:04.955 --> 00:37:09.860 IMSI catcher. So we compute score over these different events, doing stream 00:37:09.860 --> 00:37:14.830 analysis on everything that happens on your phone and eventually then come out 00:37:14.830 --> 00:37:20.820 with a warning. If the score crosses a certain threshold, there's a bunch more we 00:37:20.820 --> 00:37:25.030 would have wanted to include that's even on a Qualcomm chipset in it's debug mode 00:37:25.030 --> 00:37:29.960 not available. So this is still ongoing work as these chipsets progress and may give 00:37:29.960 --> 00:37:37.168 us more information in the future. Now, if you do find alerts, let's call them alarms 00:37:37.168 --> 00:37:41.044 on your phone. We'd be grateful if you could share them. Now, as I said, this is 00:37:41.044 --> 00:37:48.080 optional, right? You get you get the alerts shown in shown in your little tool 00:37:48.080 --> 00:37:52.730 and then you can choose to upload whichever ones you think should be shared 00:37:52.730 --> 00:37:59.697 if we get enough of them and and think that there's really hot spots of of of 00:37:59.697 --> 00:38:03.419 abuse, of course, we'll try to make that transparent, perhaps even put little dots 00:38:03.419 --> 00:38:07.950 on the GSM website so people know where abuse could be happening around 00:38:07.950 --> 00:38:20.370 demonstrations, around embassies, wherever. Applause 00:38:20.370 --> 00:38:23.410 You can also actively choose to 00:38:23.410 --> 00:38:28.090 submit data by by running an active test now usually the phone looks at everything 00:38:28.090 --> 00:38:32.370 that you produce, your phone calls, your SMS that's always stored on the phone. 00:38:32.370 --> 00:38:37.880 There's no way to upload that. And you compute a score for how secure your 00:38:37.880 --> 00:38:42.410 network is using the exact same metrics that we use on gsmmap. So that's all 00:38:42.410 --> 00:38:47.410 ported to the phone now. But if you feel like the score on gsmmap is heavily outdated, 00:38:47.410 --> 00:38:51.860 click this button. It runs some benign tests, has nothing to do with your transactions. I 00:38:51.860 --> 00:38:55.640 guess your location where you're currently connected would be included in the data 00:38:55.640 --> 00:39:02.030 and it uploads it to gsmmap. So that becomes better and better. And we can spot 00:39:02.030 --> 00:39:09.780 more networks that, for instance, like any encryption at all. Yeah, so what's what 00:39:09.780 --> 00:39:15.370 what are you what I like you to do, I think you should do to better protect 00:39:15.370 --> 00:39:20.076 yourself from mobile abuse, of course you could keep waiting for your mobile 00:39:20.076 --> 00:39:24.900 networks to fix all these issues, which I must say more recently, more networks have 00:39:24.900 --> 00:39:30.150 moved to fix issues, but still not the majority. And no network has even started 00:39:30.150 --> 00:39:35.550 to address the majority of issues. So it's just scratching the surface. So what I'd 00:39:35.550 --> 00:39:41.770 rather have you do is start defending yourself. Check out gsmmap, see if you 00:39:41.770 --> 00:39:45.800 are on a network that generally protects things like encryption. You saw the 00:39:45.800 --> 00:39:51.750 networks that lack encryption. Don't use those. And if you really choose to self 00:39:51.750 --> 00:39:58.241 defense, download, SnoopSnitch, this new tool and actively look out for abuse, for 00:39:58.241 --> 00:40:03.080 Silent SMS, binary SMS that you receive, for empty pagings, for IMSI catcher 00:40:03.080 --> 00:40:10.490 evidence and help us grow this database of abuse. Right. Also help us grow the 00:40:10.490 --> 00:40:15.720 tool base that we use. This is released open source and we put in a lot of work to 00:40:15.720 --> 00:40:20.710 make the data accessible. But now it is accessible, right? Just take it as a 00:40:20.710 --> 00:40:26.920 library and go wild with it. Do whatever you always wanted to do with raw baseband 00:40:26.920 --> 00:40:34.300 data on 2G, 3G, 4G. I am very much looking forward to your contributions to this and 00:40:34.300 --> 00:40:37.720 all that's left for me to say is thank you very much. 00:40:37.720 --> 00:40:47.570 applause 00:40:47.570 --> 00:40:57.240 Herald: Thank you, Karsten, then we will beginning with the Q&A, please, for 00:40:57.240 --> 00:41:03.590 everybody that will be asking questions, please line up on the microphones in the 00:41:03.590 --> 00:41:13.660 room and for people that exit the room, please do it with no noise and quickly. 00:41:13.660 --> 00:41:17.390 Karsten: Now, before getting into the question, let me give you one reason to 00:41:17.390 --> 00:41:22.520 actually do leave now. There's a workshop happening right now or in a few minutes 00:41:22.520 --> 00:41:27.850 that will explain how this tool works and what it can all do. We'll have an IMSI 00:41:27.850 --> 00:41:31.240 catcher there a day or so. You can tell us how that feels like being connected to an 00:41:31.240 --> 00:41:36.210 IMSI catcher. It's happening in room C, which is when you exit here one floor 00:41:36.210 --> 00:41:41.750 down and to this end. Herald: And additional information, the 00:41:41.750 --> 00:41:51.407 workshop that's Karsten says start at nineteen forty five. 00:41:51.407 --> 00:42:00.050 K: And now to your questions. distant noise 00:42:00.050 --> 00:42:04.800 K: Sure. Herald: OK, microphone number two and 00:42:04.800 --> 00:42:10.460 please, before before we before you can start number two, please do it with no 00:42:10.460 --> 00:42:19.270 noise that we hear the question from the audience. OK, number two, please. 00:42:19.270 --> 00:42:23.260 Mic 2: Thank you. Can you quickly say a few words about why it wouldn't work on 00:42:23.260 --> 00:42:27.610 custom ROMs? Because we could just install it into cyanogen phones and apparently 00:42:27.610 --> 00:42:34.750 installed and it seems to work. K: Oh, OK. So the way I understood custom 00:42:34.750 --> 00:42:38.920 ROMs is that they first remove a bunch of stuff from the phone and then put a bunch 00:42:38.920 --> 00:42:44.025 of stuff on it. Part of what we need are these proprietary Qualcomm libraries and 00:42:44.025 --> 00:42:47.050 at least on the phones where we tried cyanogen mod and what they are being 00:42:47.050 --> 00:42:51.730 removed. So if cyanogen mod could stop doing that, it would work beautifully. 00:42:51.730 --> 00:42:56.430 It's not that we need anything additional. We just need less to be deleted. 00:42:56.430 --> 00:43:04.290 Mic 2: OK, thank you. Herald: OK. Microphone number …, will you 00:43:04.290 --> 00:43:09.760 ask. OK, are there some questions from the IRC? 00:43:09.760 --> 00:43:16.090 K: I think we have a bunch of questions. Signal Angel: Actually, there is five 00:43:16.090 --> 00:43:24.030 questions, so I will just ask one or two for starting. The first one is, can all 00:43:24.030 --> 00:43:30.690 these shown attacks that you proved on your speech be mitigated by… by higher 00:43:30.690 --> 00:43:37.300 protocols levels, like encrypted VoIP or TextSecure, things like that? And what 00:43:37.300 --> 00:43:41.910 will be the residual risks? K: Mm, yeah. A good question. So how much 00:43:41.910 --> 00:43:46.740 can you protect yourself by using the mobile network less on using it as a dumb 00:43:46.740 --> 00:43:52.710 pipe, I guess is the question, what if you use just apps to call and send text? Well, 00:43:52.710 --> 00:43:59.090 obviously your calls and texts won't be intercepted anymore if they are encrypted 00:43:59.090 --> 00:44:04.560 one more time in a way that's not breakable. However, this does not solve 00:44:04.560 --> 00:44:09.100 the location tracking. It does not solve the fraud. It does not solve the denial of 00:44:09.100 --> 00:44:13.790 service. It does not solve the spamming. So you are tied to a mobile network and it 00:44:13.790 --> 00:44:18.140 has a lot of control over you, your location and your phone bill. None of that 00:44:18.140 --> 00:44:25.590 is going to go away. Herald: Another question from the IRC, one. 00:44:25.590 --> 00:44:33.380 Signal Angel: Yeah, um, the second one is: Wouldn't it be easier to design from 00:44:33.380 --> 00:44:39.902 scratch a new mobile mobile network than trying to find all flaws from actual 00:44:39.902 --> 00:44:45.080 networks, which is an endless task? K: Or I don't know where you would even 00:44:45.080 --> 00:44:49.770 start designing everything from scratch completely? The closest that I can think 00:44:49.770 --> 00:44:54.280 of designing the mobile network from scratch is LTE in the name of long term 00:44:54.280 --> 00:44:58.500 evolution. It really wants to change everything, but gives it a couple of years 00:44:58.500 --> 00:45:02.690 but as Tobias pointed out, those issues we pointed out today, they are 00:45:02.690 --> 00:45:08.220 again included in LTE. Diameter is the interconnect protocol. So we already 00:45:08.220 --> 00:45:13.410 missed a chance to to remove much of this issues by just upgrade. We'll have to fix 00:45:13.410 --> 00:45:18.950 it through firewalls and monitoring like we never got to update the Internet. 00:45:18.950 --> 00:45:22.540 Herald: OK, microphone number four, please. 00:45:22.540 --> 00:45:27.620 Mic 4: Yet just a short thing. Could you just provide a list of those libraries 00:45:27.620 --> 00:45:35.630 you need from the stock images? So I think it's pretty easy to copy them to this 00:45:35.630 --> 00:45:38.484 cyanogen mod images. K: Ok 00:45:38.484 --> 00:45:40.516 Mic 4: OK, and if the app is open source, 00:45:40.516 --> 00:45:45.900 maybe you can put it on fdroid? K: Oh absolutely. Yes. Thank you. 00:45:45.900 --> 00:45:50.970 applause Herald: The microphone number two, please. 00:45:50.970 --> 00:45:57.560 Mic 2: Got two questions, if I understood correctly, you need to be inside the 00:45:57.560 --> 00:46:02.350 operator network to actually perform those SS7 queries, right? 00:46:02.350 --> 00:46:08.030 K: Um, well, I would I would like for this to be the case. But currently, does 00:46:08.030 --> 00:46:12.020 anybody in the world connected to SS7 can send his queries. 00:46:12.020 --> 00:46:17.960 Mic 2: OK, so my question is that what was your hook point for actually doing this 00:46:17.960 --> 00:46:20.890 test? K: I think I'll quote Tobias here by 00:46:20.890 --> 00:46:23.420 saying I would rather not say anything about that. 00:46:23.420 --> 00:46:29.800 Mic 2: OK, so the second question is about the case you mentioned it's if I am not 00:46:29.800 --> 00:46:37.840 mistaken, is the session key. Right? It's and it should involve that nonce value, right? 00:46:37.840 --> 00:46:42.850 K: Yeah. Mic 2: So if it is, it already has the nonce 00:46:42.850 --> 00:46:48.130 value. So in order the attack to work, we also need to intercept the initial 00:46:48.130 --> 00:46:54.930 messages, the nonce exchange between the target and the basis station. Is that 00:46:54.930 --> 00:46:59.460 correct? K: No, the nonce is… as as they are. So 00:46:59.460 --> 00:47:05.660 the SIM card knows which key to produce. Yes. But it helps the phone to find the 00:47:05.660 --> 00:47:09.780 right encryption key. We are not the phone. We don't have the SIM card. Right. 00:47:09.780 --> 00:47:12.600 If you just give us the encryption key, we don't need the nonce. 00:47:12.600 --> 00:47:18.700 Mic 2: Yes. So what you're saying is that the query you're sending there, it 00:47:18.700 --> 00:47:25.910 actually sends you not only the encryption key, but also the nonce that is required.. 00:47:25.910 --> 00:47:30.030 K: It doesn't send us the nonce and we don't need the nonce. We can take that 00:47:30.030 --> 00:47:32.430 offline now, explain how everything works. Thank you. 00:47:32.430 --> 00:47:35.780 Herald: To microphone number three, please. 00:47:35.780 --> 00:47:40.680 Mic 3: First of all, thank you for a very good presentation and very impressive work 00:47:40.680 --> 00:47:45.330 you've done here. applause 00:47:45.330 --> 00:47:50.050 K: Thank you. Mic 3: The question I have might be a 00:47:50.050 --> 00:47:55.090 little naive, but have you also, besides taking a look at this closing this whole 00:47:55.090 --> 00:48:00.630 issue technically wise, also been taking a look into how what measures can be taken 00:48:00.630 --> 00:48:04.900 legally, at least in Germany and some countries in Europe now that we have 00:48:04.900 --> 00:48:11.431 disclosed that basically certain rules / laws have not been fulfilled, that we can 00:48:11.431 --> 00:48:15.950 enforce the operators to implement this stuff on legal ways? 00:48:15.950 --> 00:48:21.420 K: We have not looked into it. Of course, we consider the possibility as soon as 00:48:21.420 --> 00:48:25.470 somebody has an overview of where these attacks happen. And that seems to be the 00:48:25.470 --> 00:48:31.140 issue right now. There's zero attack transparency. Nobody is looking for these 00:48:31.140 --> 00:48:38.300 issues. And partly that's to the to their own disbenefit, because as soon as they do 00:48:38.300 --> 00:48:43.190 look for this issue, some of these attack patterns are very easy to stop, as I said, 00:48:43.190 --> 00:48:49.660 two German networks, mitigated them within two weeks. And these issues had been open 00:48:49.660 --> 00:48:54.510 for 20 years. Had they ever looked into their own data, that would have seen this 00:48:54.510 --> 00:49:00.060 going on. So I'm not very confident that anybody in Germany at least has an 00:49:00.060 --> 00:49:04.650 overview of where abuse would come from. And as soon as it does, I don't think 00:49:04.650 --> 00:49:10.310 there's much point in litigating. Let's just stop the possibility of abuse. Right, 00:49:10.310 --> 00:49:14.990 instead of complaining about it happening. But I'm with you. If there's corner cases 00:49:14.990 --> 00:49:19.660 in which abuse just can't be stopped, let's fight it legally, of course. Right. 00:49:19.660 --> 00:49:24.850 And if all of you contribute information through SnoopSearch, does the empty 00:49:24.850 --> 00:49:29.560 pagings, if we can find patterns of abuse, of course, we'll aggregate them and 00:49:29.560 --> 00:49:36.680 try to move against them. Herald: OK, microphone number four, 00:49:36.680 --> 00:49:40.740 please. Mic 4: You said you can buy your way into 00:49:40.740 --> 00:49:46.790 the SS7 Network, but how easy is it actually to get your access? And what do 00:49:46.790 --> 00:49:50.690 you estimate: How many players are there in the network? Can you give any 00:49:50.690 --> 00:49:54.311 estimation? K: I have absolutely no idea. I know that 00:49:54.311 --> 00:50:01.760 there's some 800 companies who who are legally allowed to access SS7 and then 00:50:01.760 --> 00:50:06.860 those, of course, have subcontractors, legal and illegal, and some people who 00:50:06.860 --> 00:50:11.186 bribe them. Yet other people who hack their systems or the systems of the 00:50:11.186 --> 00:50:14.920 subcontractors, it's very hard to estimate. No idea. But definitely too many 00:50:14.920 --> 00:50:18.650 to trust all of them. Mic 4: And would it be possible for me to 00:50:18.650 --> 00:50:25.710 get access to this without any operator stuff or. I don't want to operate a phone 00:50:25.710 --> 00:50:31.300 network, but I want to have access because I want to provide a service, some service? 00:50:31.300 --> 00:50:35.670 K: Well, I wish the answer was no, but of course, right of to be as an I and a bunch 00:50:35.670 --> 00:50:40.910 of other people can get access. You should be able to get that too. But I'm not going 00:50:40.910 --> 00:50:44.600 to tell you how. laughter and applause 00:50:44.600 --> 00:50:51.680 Herald: Yet another question from the IRC. Signal Angel: We're about nine questions, 00:50:51.680 --> 00:50:58.200 so no problem for me. First one, what about Windows phones, jail breaked 00:50:58.200 --> 00:51:04.890 iPhones, or something like this will the app in the end [be] on this phones? 00:51:04.890 --> 00:51:11.250 K: Our app doesn't run on anything other than Android, but the chipsets are, of 00:51:11.250 --> 00:51:16.670 course, the same. So if you can speak to a chipset through a jail broken iPhone, for 00:51:16.670 --> 00:51:22.070 instance, you could create a similar application. We just wanted to target the 00:51:22.070 --> 00:51:25.990 biggest population of phones, and that seems to be Android phones. 00:51:25.990 --> 00:51:33.160 Herald: Then number two, please. Mic 2: One further thought on self-defense 00:51:33.160 --> 00:51:41.110 as self-defense has don't has to be proportionate, I think, and identities are 00:51:41.110 --> 00:51:46.771 not secure in the digital sphere. How about developing some proactive, as we 00:51:46.771 --> 00:51:52.820 heard the word defense tools? K: Proactive as in hack the networks, 00:51:52.820 --> 00:51:59.010 until they have no chance but to fix? Mic 2: That's what you understood, but. 00:51:59.010 --> 00:52:03.010 But, I support that. laughter K: I'm not going to say that I dislike the 00:52:03.010 --> 00:52:07.620 idea. But you won't see me here next year explaining how I did it. 00:52:07.620 --> 00:52:11.690 Mic 2: Thank you. Herald: Microphone number three, please. 00:52:11.690 --> 00:52:17.070 OK. When did you check the other two German networks didn't fix the identifier 00:52:17.070 --> 00:52:21.800 and the issue. K. Which network do you work for? 00:52:21.800 --> 00:52:27.780 Mic 2: I'm Holger. We talked last week. K: Yeah. So yeah. Maybe you fixed it too. 00:52:27.780 --> 00:52:30.930 We didn't, we didn't check. Mic 2: We fixed it within 24 hour, 24 00:52:30.930 --> 00:52:34.590 hours after our call. K: Wow. OK. 00:52:34.590 --> 00:52:38.300 Mic 2: On both networks. applause 00:52:38.300 --> 00:52:44.430 Thank you. Better late than never. Thank you. 00:52:44.430 --> 00:52:47.320 Mic 2: That's right. K: OK, so that's three out of four now, 00:52:47.320 --> 00:52:52.610 that fix one out of 100 problems. Mic 2: No, it's… I know that's why we 00:52:52.610 --> 00:52:59.610 don't go to the press and don't tell that SS7 is fixed and we know we still have 00:52:59.610 --> 00:53:06.920 problems also. It's all four. I work for Telefonica, which is O2 and eplus. 00:53:06.920 --> 00:53:11.291 K: Oh yeah. Well, congratulations. Sorry. Sorry for spoiling your Christmas. 00:53:11.291 --> 00:53:13.440 laughter 00:53:13.440 --> 00:53:19.400 Herald: Microphone number two, please. Mic 2: I'd like to know why these empty 00:53:19.400 --> 00:53:24.180 pagings occur in the context of the location tracking, I thought, as soon as 00:53:24.180 --> 00:53:30.620 the phone registers in the network, the base station, which is this connected to, 00:53:30.620 --> 00:53:32.630 is known in the network anyway. Is that the case? 00:53:32.630 --> 00:53:37.490 K: That's a very good question. And let me let me go back to one earlier slide to to 00:53:37.490 --> 00:53:45.590 explain that, one second, so that the empty pagings do not occure when you send 00:53:45.590 --> 00:53:50.380 these creepy AnytimeInterrogation messages. They are just there for spying 00:53:50.380 --> 00:53:55.280 and there's no way to page the customer. But since this got blocked and Tobias went 00:53:55.280 --> 00:53:59.070 into great level of detail explaining this, you need a couple of other messages 00:53:59.070 --> 00:54:03.320 to now track some of this location and these messages when meant for location 00:54:03.320 --> 00:54:09.530 tracking them and ment for other purposes. For instance, as I provide subscriber info 00:54:09.530 --> 00:54:14.950 that however you reach it is always the last message you need. This does do a 00:54:14.950 --> 00:54:19.020 paging and then to provide subscriber info really makes no sense unless you send 00:54:19.020 --> 00:54:23.890 something afterwards also, deliver an SMS connect to call or whatever. So the paging 00:54:23.890 --> 00:54:29.690 is already sent in anticipation that an SMS will come or that the call will come. 00:54:29.690 --> 00:54:33.880 But if you're only the creepy guy tracking it, they're going to send it SMS and 00:54:33.880 --> 00:54:38.410 that's where the empty paging comes from. Mic 2: OK, but still also in these cases 00:54:38.410 --> 00:54:43.610 where something follows the paging, isn't it a type of double checking whether it's 00:54:43.610 --> 00:54:50.230 really there or I mean, the location info itself should already be present and the 00:54:50.230 --> 00:54:53.510 network, isn't it? K: Yeah, yeah. It just reconfirms that the 00:54:53.510 --> 00:54:57.640 subscriber is really there. So it's basically saying: Somebody you just 00:54:57.640 --> 00:55:01.370 interrogated your location because they want to send you something. Let's check 00:55:01.370 --> 00:55:05.350 that you're really still there because otherwise we'll tell them something wrong. 00:55:05.350 --> 00:55:10.420 But Tobias do you want to comment on that. Tobias: Yeah. OK, so the empty paging is 00:55:10.420 --> 00:55:15.930 not anticipation or something that's coming after. It's to get the current cell 00:55:15.930 --> 00:55:20.970 that you are located at, because when you are moving around in your location area 00:55:20.970 --> 00:55:24.850 and the area that is covered by the switching center that you're currently 00:55:24.850 --> 00:55:31.120 being served by, your phone doesn't necessarily contact the base station. So 00:55:31.120 --> 00:55:37.790 it could be that that the networks last position of you is somewhere you received 00:55:37.790 --> 00:55:43.950 an SMS or text or call, and then you moved to a completely different area if your 00:55:43.950 --> 00:55:49.130 phone didn't have network contact in the meantime, the network would still only 00:55:49.130 --> 00:55:55.610 know the last point of contact. So that's why the why the empty paging happens so 00:55:55.610 --> 00:56:01.310 that the that the network knows the base station that's actually currently closest 00:56:01.310 --> 00:56:06.780 to you. That's also why the law enforcement uses a lot of Silent SMS so 00:56:06.780 --> 00:56:12.530 that that they can get the last position in the network. And it's also an option if 00:56:12.530 --> 00:56:17.240 you send provide subscriber information, you can just send it and get back the last 00:56:17.240 --> 00:56:23.720 known position without a paging or you can set the current location flag and provide 00:56:23.720 --> 00:56:29.860 subscriber information. And only then the subscriber gets paged and you will receive 00:56:29.860 --> 00:56:33.530 the current location. K: And that's that's one good example for 00:56:33.530 --> 00:56:37.880 how SS7, which is supposed to be so insecure we can never fix it, can 00:56:37.880 --> 00:56:42.750 easily be fixed. There's an option that says we're using this as normal feature 00:56:42.750 --> 00:56:46.480 that's absolutely needed. And we have this creepy extension to also ask for the 00:56:46.480 --> 00:56:51.140 location. And some networks choose to not answer that. The answer was zero zero zero 00:56:51.140 --> 00:56:57.540 zero and nothing broke. Right. So you can just ignore the insecure parts of SS7 and 00:56:57.540 --> 00:57:01.890 do whatever you think is right. And for the most part, it continues to work. But 00:57:01.890 --> 00:57:04.040 I think we're well beyond answering your question now right? 00:57:04.040 --> 00:57:11.230 Mic 2: No, but from your answers. Thank you very much. But another question 00:57:11.230 --> 00:57:16.710 arises, because if it's actually to locate your phone and to find out which cell 00:57:16.710 --> 00:57:23.310 you're actually in, then it implies that it's not only one base station that since 00:57:23.310 --> 00:57:29.190 the paging call, but a whole bunch of base stations. Do you know something about the 00:57:29.190 --> 00:57:35.260 algorithm? I mean, how many around the last known location are paging everybody 00:57:35.260 --> 00:57:39.560 nationwide or how does.. K: Everybody can implement this as they 00:57:39.560 --> 00:57:45.340 wish? And I don't have much insights into how 3G does it, but in 2G typically is: 00:57:45.340 --> 00:57:49.730 There's one paging send in the last cell that saw you. You don't respond. It's send 00:57:49.730 --> 00:57:53.600 in a larger area. You don't respond. It's sent for the whole location area. And then 00:57:53.600 --> 00:57:58.100 some networks, you don't respond. They send it in the entire country. But that's 00:57:58.100 --> 00:58:01.589 rare. Right? Mic 2: Thank you very much. 00:58:01.589 --> 00:58:12.790 Herald: Okay. Questions from the IRC? Signal Angel: Did SnoopSnitch allow you to 00:58:12.790 --> 00:58:20.740 reveal any kind of attack in countries. Not special name in mind. 00:58:20.740 --> 00:58:26.920 K: Does it allow you to detect attacks in countries? Yeah, yeah, some kind of 00:58:26.920 --> 00:58:32.520 Tapsell. I think the answer is yes. Its whole purpose is to detect attacks. And it 00:58:32.520 --> 00:58:35.852 also works in countries… laughter 00:58:35.852 --> 00:58:39.840 Herald: Did you succeed in detecting attacks. K: Did we succeed in 00:58:39.840 --> 00:58:46.590 detecting. Yes, we did. And if you go down to the Saal C, Room C, you can see how it's 00:58:46.590 --> 00:58:53.880 currently people are being attacked and currently they detect that. Ok 00:58:53.880 --> 00:58:59.280 Herald: OK microphone number five, please. Mic 5: Yes, thanks, it's going back to SS7 00:58:59.280 --> 00:59:05.670 basics. Can you quickly explain how SS7 is implemented? Is this a VPN on the public 00:59:05.670 --> 00:59:10.610 Internet through the providers? What's the technical reality of transport? 00:59:10.610 --> 00:59:16.640 K: That's a very good question. Of course, that's a very good question. And I only 00:59:16.640 --> 00:59:21.890 have half of the information, too. I keep learning. But so it seems that it was 00:59:21.890 --> 00:59:27.430 implemented initially as a network between Western European telcos and their run 00:59:27.430 --> 00:59:33.961 cables, dedicated cables for SS7. SIGTRAN they called this and then a couple 00:59:33.961 --> 00:59:38.250 more networks connected to it. And each of them had to run the cable to one of the 00:59:38.250 --> 00:59:42.690 other telcos. But eventually they changed that and then introduced what I call 00:59:42.690 --> 00:59:46.741 routing providers. So telcos are not connected to each other usually, but 00:59:46.741 --> 00:59:52.240 through a routing provider like on the Internet and those routing providers, they 00:59:52.240 --> 00:59:56.710 typically don't run a cable to your house anymore. If you are a new telco, they give 00:59:56.710 --> 01:00:00.790 you a VPN over the Internet. So it's diverse. I'm sure there's still some 01:00:00.790 --> 01:00:04.790 dedicated lines between Germany and France, say, and there's some others 01:00:04.790 --> 01:00:08.510 connecting and these big clouds that are routing providers. And it's actually 01:00:08.510 --> 01:00:12.290 really difficult to get your address routed everywhere in the world. So even if 01:00:12.290 --> 01:00:16.886 you connect to SS7, all you're connected to is one routing provider and that 01:00:16.886 --> 01:00:21.690 routing provider knows that you own these addresses. Now it's up to you to convince 01:00:21.690 --> 01:00:25.850 every other of the big seven or nine, depending on how you count routing 01:00:25.850 --> 01:00:34.250 providers that you are that guy with those addresses. So the BGP equivalent of SS7 is 01:00:34.250 --> 01:00:40.410 to get nine roaming agreements signed with people on these other nine operators and 01:00:40.410 --> 01:00:44.810 then fax those roaming agreements to everybody else involved. So they type it 01:00:44.810 --> 01:00:49.530 into your computer, into their computers, very manual and very hard to grow the 01:00:49.530 --> 01:00:52.830 network. But for the most part, it doesn't change, of course- 01:00:52.830 --> 01:00:57.940 Mic 5: So that the low level transport is not really an attack surface from the 01:00:57.940 --> 01:01:00.840 public Internet. K: It can be the low level transport can 01:01:00.840 --> 01:01:07.090 be an attack surface if people just stupidly leave open their local networks. 01:01:07.090 --> 01:01:11.156 But it's rare. It's much more common, speaking about our talk next year, 01:01:11.156 --> 01:01:15.844 hopefully on the other interconnect networks, there's one interconnect network 01:01:15.844 --> 01:01:22.240 for data roaming. It's called GRX. And since everything is IP anyway on data 01:01:22.240 --> 01:01:26.610 roaming, people sometimes do leave it out on the Internet or just do it unencrypted 01:01:26.610 --> 01:01:31.010 over the Internet. And it does seem to become more popular also with the SS7 01:01:31.010 --> 01:01:37.440 replacement Diameter, which again is pure IP. So there's no dedicated thing that you 01:01:37.440 --> 01:01:41.660 first have to encapsulate in a VPN before you can route it over the Internet. You 01:01:41.660 --> 01:01:47.060 can run Diameter over the open Internet if you want. It's stupid, but people seem to 01:01:47.060 --> 01:01:52.170 do it anyway. Herald: OK, the microphone number six, 01:01:52.170 --> 01:01:55.310 please. Mic 6: OK, my question is, if you could 01:01:55.310 --> 01:02:00.451 comment why these message were put in the protocol at the first place, it they are 01:02:00.451 --> 01:02:07.270 so easy to block and to fix. And the other question is, if all the other problems 01:02:07.270 --> 01:02:11.620 that you pointed out are as easy to fix for the network operators. 01:02:11.620 --> 01:02:16.780 K: So I don't have an answer to your first question. Why do you put a tracking 01:02:16.780 --> 01:02:22.470 message in the standard and then call it AnytimeInterrogation, gosh, like that 01:02:22.470 --> 01:02:25.610 invokes feelings for me, interrogation room and all. I mean, this 01:02:25.610 --> 01:02:30.440 is spy stuff, right? And there's no practical, purposeful but. Right. Who 01:02:30.440 --> 01:02:35.000 wrote SS7 standard? Western European governments being afraid of the Russians, 01:02:35.000 --> 01:02:39.060 of their own citizens, who knows? Right. I don't know why they put every single 01:02:39.060 --> 01:02:44.280 message in, though. So your second question was what again? 01:02:44.280 --> 01:02:49.060 Mic 6: If the other vulnerabilities are as easy as to fix? Or just blocking messages. 01:02:49.060 --> 01:02:55.730 K: No they're not. And I tried to point that out in one of the slides that… that 01:02:55.730 --> 01:03:02.270 AnytimeInterrogation can be fixed, as can, for instance, as does SendIdentification 01:03:02.270 --> 01:03:07.310 message, right. You just block that has no purpose, routing this internationally. But 01:03:07.310 --> 01:03:11.600 the other queries on this page, at least you need those internationally, at least 01:03:11.600 --> 01:03:17.430 to enable roaming. So the best you can do is, as I said, first block these queries 01:03:17.430 --> 01:03:21.010 from anybody who's not your roaming partner, right? Don't respond to those 01:03:21.010 --> 01:03:26.520 people and then do some plausibility checking, secondly, make sure that if a 01:03:26.520 --> 01:03:31.380 subscriber is actually in your own network, that you don't honor requests from another 01:03:31.380 --> 01:03:36.600 country. Right. And that should remove most of the issues because most abuse comes from 01:03:36.600 --> 01:03:40.340 other countries. It's just more likely if there's 800 parties connected to this 01:03:40.340 --> 01:03:46.901 network that the one doing the abuse is not yours. Good question. Thanks. 01:03:46.901 --> 01:03:59.000 Subtitles created by c3subtitles.de in the year 2021. Join, and help us!