WEBVTT 00:00.790 --> 00:06.590 Okay, welcome to my talk, COVID-1984, propaganda and surveillance during a pandemic. 00:08.090 --> 00:13.830 Before we start, I would like to make a brief introduction both to this talk and to myself. 00:14.550 --> 00:32.610 My name is Mauro Eldritch, I work as a cyber security architect, I'm the founder of DC5411 Argentina, and I was a speaker for DEF CON Las Vegas, Dragonjar Colombia, POSCON Iran, and Texas Cyber Summit, among other conferences. 00:32.610 --> 00:38.270 I'm a returning speaker to this village, I spoke here on DEF CON 26. 00:40.430 --> 00:53.510 And this talk is about explaining the political situation of surveillance and propaganda in Argentina, although it is also compatible with the same situation in other countries and regions. 00:53.510 --> 01:08.930 This will be explained from a hacker's point of view, using a hacker's toolbox, which in my case includes social engineering and open source intelligence, and there's even some application reverse engineering at the end. 01:10.270 --> 01:14.590 This talk will be divided into two chapters. 01:14.610 --> 01:22.170 The first propaganda about the Argentine pro-government propaganda apparatus and social networks seen from the inside. 01:22.170 --> 01:31.950 We will see the process from the scratch, from the infiltration of a sock puppet account into the apparatus to its internal operation. 01:32.890 --> 01:36.250 The second part is surveillance. 01:36.310 --> 01:44.750 We will dissect the Cuidar COVID-19 mobile application, which is now mandatory by law in order to circulate. 01:44.750 --> 01:56.130 So if you want to leave your home, if you want to step away from your doorstep, you have to have this application installed on your phone. 01:57.090 --> 02:11.390 And we will dissect this application uncovering many privacy abuses, bad practices, and lots of material worth of appearing on Reddit, especially on Programming Horror subreddit. 02:12.510 --> 02:20.370 Just a brief disclaimer, every item disclosed here is publicly available through OSINT. 02:20.370 --> 02:35.810 On my jurisdiction, it is totally legal to reverse engineer any application on any software, and I wasn't involved in any illegal activity directly or indirectly. 02:36.030 --> 02:41.870 So that being said, let's start with our first chapter, propaganda. 02:44.290 --> 02:48.890 So, what is propaganda, you might ask? 02:49.090 --> 03:07.330 Propaganda is a communication that is primarily focused on influencing an audience and furthering an agenda, which might not be objective or might not be real at all, and that may be presenting facts selectively. 03:07.330 --> 03:13.870 You might have heard the term alternative facts recently in media. 03:13.870 --> 03:18.590 Well, it's probably related to this kind of behavior. 03:20.390 --> 03:24.150 Online, agents of propaganda are called trolls. 03:24.630 --> 03:33.710 They massively comment on social networks, supporting a certain government, in this case, or a certain movement. 03:33.710 --> 03:38.450 It may not always be a government, but a political movement. 03:39.290 --> 03:42.970 They try to establish their own debates and trends. 03:42.970 --> 03:52.140 For instance, in my country, it's pretty popular to see trendings like thanks, President Fernandez, or we support you, President Fernandez. 03:53.530 --> 03:59.950 They are also maybe found diverting the focus from opposition debates. 03:59.950 --> 04:28.870 For example, every time an opposition part tries to open a thread or tries to communicate something, you might find the trolls leaving spam comments, fancams, or any other material that, when exposed repeatedly, could turn this thread into something really difficult to follow for legit users or for interested users. 04:29.130 --> 04:34.850 So it may turn the conversation into something unbearable. 04:35.900 --> 04:49.700 They also work by establishing negative trends against the opposition, and they are usually grouped into troll farms, where they work together in an organized manner, surprisingly. 04:51.700 --> 04:54.380 How do you recognize a troll? 04:54.960 --> 05:05.040 Okay, first, it's pretty simple, because lots of default configurations on their profile are present. 05:05.080 --> 05:09.400 For example, they don't have a picture, a profile picture. 05:09.720 --> 05:13.400 They use a fake one or a stock one. 05:13.840 --> 05:16.320 There are lots of numbers in their handlers. 05:16.320 --> 05:20.000 For instance, Mauro and a bunch of numbers. 05:20.060 --> 05:30.280 That's pretty common, pretty normal on Twitter, actually, because that's the default username that Twitter gives you and allows you to change later. 05:31.440 --> 05:33.960 They share common terms and phrases. 05:33.960 --> 05:37.240 They have their own languages, surprisingly. 05:37.660 --> 05:40.960 They agree on a common version to answer debates. 05:40.960 --> 05:45.620 For example, the president's offshore accounts are a media operation. 05:46.200 --> 05:52.260 So they are basically saying that everyone else is lying, and they are not. 05:52.900 --> 05:55.880 And they obviously work in swarm behavior. 05:55.880 --> 05:58.300 They never come alone. 05:58.740 --> 06:01.100 They act like a swarm. 06:02.060 --> 06:07.920 Many governments are credited with having online propaganda apparatus and agents. 06:07.920 --> 06:13.380 For example, Russia, with the collapse of the Trolls of Olgino. 06:13.380 --> 06:16.320 China and the 57th party. 06:16.400 --> 06:19.660 North Korea, with the United Front Department. 06:19.680 --> 06:28.020 Venezuela, with the Armada Bolivariana de Trolls, which barely translated to English is Bolivarian Army of Trolls. 06:28.020 --> 06:43.340 And also Argentina, with the Ciber-Cas, or Tropa-Ca, which translated from Spanish is K-Troop or Cyber-K, which is not actually a K-pop supporting fandom. 06:44.040 --> 06:45.720 It's valid, the clarification. 06:47.620 --> 06:56.860 This global advance on propaganda apparatus around the world is such that there are social networks with firm positions towards them. 06:57.340 --> 07:09.460 For example, Counter-Social is against it, and countries with the highest propaganda incidents are absolutely banned, blocked entirely by their IP block. 07:09.460 --> 07:11.630 Some of them are listed here. 07:13.580 --> 07:23.220 And there are other social media sites, for example, this one in my country, that supports the propaganda behavior. 07:23.440 --> 07:34.980 Facebookular.net was an Argentine social network against establishment and imperialism, and it's a firm supporter of the Peronist party. 07:35.380 --> 07:43.540 It was a place where only its militants participated, and it was backed, supported, by the Ministry of Culture. 07:44.400 --> 07:55.480 At a first sight, you might obviously see that this is a bad rip-off of Facebook, and it even allows you to register with Facebook. 07:55.860 --> 08:09.700 This project is now defunct, and it's no longer active, but it was heavily used by users from Argentina and Venezuela some years ago. 08:12.220 --> 08:13.820 Trolls and censorship. 08:14.220 --> 08:20.740 The goal of a troll is not only to spread messages, but to prevent the opposition from doing so. 08:21.240 --> 08:28.940 Lots of groups participate in massively reporting legit posts or accounts from the opposition in order to take them down. 08:29.620 --> 08:40.240 They are actively abusing an automatic mechanism for reporting, so if a lot of users report someone, it must be true. 08:40.400 --> 08:47.640 And this automatic behavior takes down the post or the account without any real intervention. 08:49.480 --> 09:00.180 Argentina, it's an ideal terrain for this surveillance experiments, since many times the government have tried to control social networks to no avail. 09:01.460 --> 09:14.860 For example, Peronist legislators proposing that users who comment on websites identify themselves with their national identification number, which is like the social security number for US residents. 09:16.100 --> 09:24.140 Or a senator from the same party proposing a public and democratic regulation of social networks. 09:24.660 --> 09:39.560 Or the cyber patrol protocol imposed this year, which basically ended up with a lot of people being detained for tweeting or for expressing themselves on social media. 09:40.040 --> 09:50.660 You know, there are a lot of projects and already installed measures to control social networks, but what about the trolls? 09:50.660 --> 09:53.980 The trolls continue working together and growing. 09:55.260 --> 09:59.120 So this is where we start after this introduction. 09:59.120 --> 10:00.500 Tracking trolls. 10:02.200 --> 10:12.880 The Argentine apparatus has the particularity of being made up of real users as well as trolls and bots, as you can see when exploring their hashtags. 10:14.260 --> 10:19.820 Rumors indicated that many users received invitations to join this apparatus. 10:20.380 --> 10:23.940 My goal was to get that invitation. 10:24.040 --> 10:27.180 You know, it's like a golden ticket to the chocolate factory. 10:27.820 --> 10:36.000 So after a month of silent observation, I created a sock puppet account mimicking the behavior of these users. 10:37.480 --> 10:40.760 And it was configured using these settings. 10:40.760 --> 10:46.180 Profile photo, President Fernández and Vice President Fernández together. 10:46.880 --> 10:52.960 The cover photo was Vice President Fernández giving a speech at Vélez Stadium. 10:53.840 --> 10:56.240 And who was I following? 10:56.240 --> 11:03.820 The president and some ministers, along with a few real pro-government accounts. 11:04.400 --> 11:07.640 What about the description and tweets? 11:08.160 --> 11:15.780 Both I made them imitating the specific languages and symbolism used by these users. 11:15.860 --> 11:17.320 How? 11:17.460 --> 11:23.480 Well, using Twitter API to monitor their hashtags and extracts the most repeated words. 11:23.480 --> 11:27.940 You may find them here as a word cloud. 11:28.100 --> 11:35.700 Most of these are pejorative terms to refer to opponents like Bolsonaro, the former president of Brazil, Bolsonaro. 11:36.280 --> 11:41.100 And Piranha instead of Sebastián Piñera, the Chilean president. 11:42.240 --> 11:49.760 So after learning how they speak, now that we speak the same languages, it was a matter of time before getting the invitation. 11:52.400 --> 11:58.420 This sub-puppet raised a hundred or so followers in a few hours. 11:58.420 --> 12:08.140 I have my original Twitter account since 2015 and I think I have less than 300 users, just for comparing. 12:08.480 --> 12:12.540 Most of these were trolls, but a few real users too. 12:12.860 --> 12:19.260 So I started retweeting official accounts and these real users and tweeting with their same hashtags. 12:19.520 --> 12:24.620 For instance, Macri should be in jail, our ex-president from the opposition. 12:25.240 --> 12:32.620 And after only three days, I received my invitation, this golden ticket from a user we'll call DC. 12:34.700 --> 12:44.320 On the left, you can see the original conversation in Spanish and I translated it the best I could to English. 12:44.320 --> 12:47.280 Hey buddy, we are making Twitter groups to install hashtags. 12:47.300 --> 12:48.540 Wanna join? 12:49.000 --> 12:50.120 Hey buddy, sure. 12:50.120 --> 12:51.740 What I'm supposed to do? 12:51.820 --> 12:56.260 Well, basically you have to tweet the hashtags that we send to the group. 12:56.840 --> 13:06.340 After agreeing this, I was added to the group number 300 and some of soldiers of the national project containing 50 people. 13:08.320 --> 13:12.520 One of the messages shared there was this. 13:12.520 --> 13:17.560 Guys, Ariel Garbarz is asking us to use LARRETAISRESPONSIBLE. 13:18.460 --> 13:21.940 LARRETA is the mayor of Buenos Aires city and a member of the opposition. 13:22.520 --> 13:25.640 Now you may ask, who is Ariel Garbarz, by the way? 13:26.560 --> 13:28.500 No clue so far. 13:29.680 --> 13:45.940 But in fact, both of these hashtags that made the top five in my country, one of them being PESCADOPODRIDOEME and LARRETAISRESPONSIBLE are fabricated trends proposed on propaganda groups. 13:46.600 --> 13:49.600 You may find the source on getthytrends.com. 13:49.600 --> 13:57.500 Just for confirmation, I went to trending earlier and I found that this information was right. 13:57.940 --> 14:08.020 And these trendings lasted for at least 12 hours, which gives us a lead on how toxic are these propaganda interventions. 14:10.540 --> 14:26.040 Another message on this group is, hey guys, with Ariel Garbarz, we formed a group where he's the administrator and tells us what to publish and at what time, so we can get our trend to always be in first place. 14:26.460 --> 14:33.560 Another user replies, it's like you say, we used to do the same long ago with the K youth, a new group. 14:33.980 --> 14:37.700 You have to set a day and a time, it's the best. 14:38.020 --> 14:40.520 Again, who is Ariel Garbarz? 14:42.360 --> 14:46.680 Okay, remember DC, the original member who invited me here? 14:46.840 --> 14:52.260 Well, 10 days later, another account from him tried to recruit me again. 14:52.300 --> 14:55.760 So he has probably forgot about my account. 14:55.920 --> 14:58.920 Hey buddy, we're making Twitter groups to install hashtags. 14:58.920 --> 15:00.020 Wanna join? 15:00.020 --> 15:02.420 And I say, hey buddy, sure, go ahead. 15:02.980 --> 15:09.220 I was added once again to another Soldiers of the National Project 300 and something. 15:10.160 --> 15:15.680 Now by accident, I was a member of two propaganda groups. 15:16.160 --> 15:23.940 There was a six digit difference between each group ID, for example, 301 and 307. 15:24.040 --> 15:32.040 Each group can have up to 50 users, 49 if you do not take into account administrator. 15:34.060 --> 15:44.440 So 49 users multiplied by six new groups are almost 300 new users in 10 days. 15:44.440 --> 15:52.480 This is how fast can this apparatus grow without almost no any effort. 15:54.260 --> 16:05.340 Currently, according to the last group ID I managed to find, there are at least about 350 propaganda groups on Twitter alone. 16:05.340 --> 16:12.320 This gives us a total of 17,150 users. 16:13.500 --> 16:20.400 I imagine all these people, try to picture all these people tweeting five or six times a day. 16:20.400 --> 16:23.080 They can make any trend they want. 16:24.960 --> 16:29.260 Well, also, this doesn't end here. 16:29.700 --> 16:33.120 A WhatsApp group is shared with both groups I belong to. 16:33.120 --> 16:36.160 And of course, hashtags are shared there too. 16:37.360 --> 16:40.620 This group is city seven. 16:40.620 --> 16:46.220 There are seven groups of around 256 users. 16:46.940 --> 17:02.820 If we take into account 255 users, we are not taking into account administration, multiplied by these seven groups, there are almost 2,000 new users here that may be repeated. 17:04.820 --> 17:09.620 And adding them all together, we have almost 20,000 users. 17:09.620 --> 17:11.960 Again, there may be repeated users. 17:12.920 --> 17:14.800 This doesn't end here. 17:15.660 --> 17:23.400 This group had members from the USA, Spain, the United Arab Emirates, and even Germany. 17:24.140 --> 17:28.300 So, upon joining, the administrator started sending hashtags. 17:28.340 --> 17:33.620 In this case, at half past seven, we come out with $Witness. 17:33.960 --> 17:40.400 That administrator, as you may see in the message, or the leading message, is Ariel Garbaros. 17:40.520 --> 17:44.400 And that name now definitely rings a bell for you, right? 17:45.320 --> 17:50.460 So, okay, let's try to answer who is Ariel Garbaros. 17:52.260 --> 18:03.760 He's the CEO of Protección Digital, which is digital protection, a company which was favored with many contracts by the federal justice and various Argentine governments. 18:03.760 --> 18:09.820 He was benefited to direct several state infrastructure projects. 18:10.080 --> 18:15.290 He was appointed computer attorney general for the 17 and 19 elections. 18:15.520 --> 18:22.900 And he's the leader of the propaganda apparatus on social networks, the lord of the trolls, we might say. 18:22.900 --> 18:28.830 The source of this is the newspaper Perfil, where he answered an interview. 18:32.600 --> 18:39.680 Here is a video of Ariel Garbaros instructing the trolls live, but it is in Spanish. 18:39.680 --> 18:48.660 And for time reasons, I'm not going to show it here, but it will be shared on the GitHub repository along with the slides. 18:49.660 --> 18:55.020 Obviously, since in Argentina, we speak Spanish, the video is in Spanish. 18:56.260 --> 18:59.340 This was leaked on Twitter by a user. 19:00.880 --> 19:08.760 So, the final diagram of the trolls, or a simplified one, is here. 19:08.900 --> 19:29.520 Ariel Garbaros acts as a leader, sends the trending topics to be installed to its coordinators, the coordinators of each group, which then share this objective to both Twitter groups and WhatsApp groups, which then are executed by the trolls. 19:32.060 --> 19:34.500 So, is this illegal? 19:34.860 --> 19:38.460 In Twitter platform, it is against the terms of service. 19:38.660 --> 19:43.720 The Twitter safety team actively tracks and takes down state-linked propaganda groups. 19:43.860 --> 19:47.000 In our country, well, it is not. 19:47.000 --> 19:56.180 In fact, it looks suspiciously endorsed, like the Facebook ripoff social media I showed you some slides ago. 19:57.780 --> 20:00.500 What tools did you use for this chapter? 20:00.500 --> 20:16.080 Well, Trendinalia, Trends24in, Votometer, GetDayTrends, the Twitter API, the Twitter RubyGem, and my own tool, which is also in the repository, venator.lua, for recognizing bot or troll-like behavior. 20:17.400 --> 20:21.460 So, what is the propaganda apparatus up to right now? 20:21.500 --> 20:33.920 Generating hashtags supporting the use of the government application to monitor the coronavirus outbreak, you know, the Cuidar COVID-19 application, which leads us to the next part. 20:35.100 --> 20:40.360 Surveillance, or dissecting the ARGO coronavirus application. 20:41.820 --> 20:47.480 So, this is the tracking application proposed by the Argentine Ministry of Modernization. 20:47.700 --> 20:58.560 It is mandatory by law to circulate, so if you want to leave your home, you want to step outside your front door, you need to have this application installed. 20:59.320 --> 21:13.420 The current version is 3.3.1, but here we analyzed 1.0.2, 3.0.7, and 3.3, using the most common tools available for Android, the compiling and reverse engineering. 21:14.400 --> 21:31.060 At first glance, I noticed a lot of broken functions, and reviewing the code, I found many dyslexic errors, like for example, disabilidad instead of disabilidad, which in English means disabled. 21:31.060 --> 21:34.480 Bonotus instead of botones, buttons. 21:34.600 --> 21:40.220 And this problem is repeated regardless of the compiler user. 21:41.840 --> 21:44.600 Numeor instead of numero, number. 21:45.480 --> 21:48.180 Masculio instead of masculino, male. 21:50.240 --> 21:57.260 Autoevaluacoin instead of autoevaluación, which is a critical feature of this application, the auto-evaluation module. 21:59.040 --> 22:06.980 And after digging a little bit further, I found a new rolegride-only token discloser, which is not a vulnerability per se, but will become handy later. 22:09.060 --> 22:19.680 Insecure JSON structure creation, vulnerable to injection or manipulation, instead of using the Java-provided functions to build JSON. 22:20.780 --> 22:24.660 A reference to a long-dead product, Google Plus. 22:25.880 --> 22:35.740 Lots of insecure and unsanitized execute calls, really, a lot of them, and some even could lead to SQL injection. 22:36.740 --> 22:45.560 The application communicates with foreign servers, which is by national law or local law, it's not allowed. 22:47.060 --> 22:53.380 These servers are not pretty much safe for storing medical data or, well, anything. 22:54.520 --> 22:58.120 This is the map of the application. 22:58.220 --> 23:06.100 All of these assets inherit the qualification of C, so it's not secure at all. 23:06.840 --> 23:10.540 A lot of missing security features. 23:12.180 --> 23:15.520 And privacy is also at risk. 23:15.660 --> 23:24.020 Of a total score of 100, this application received a 61 score. 23:24.840 --> 23:34.200 There are many write operations that record PII about the device, which are in Base64. 23:34.500 --> 23:47.000 Whenever you decode them, you get that this application is trying to keep track of the device ID, the build of the own application, and the device manufacturer. 23:47.660 --> 23:55.700 Obviously, since this is an emulated device, the application was not able to extract any valid information at all. 23:56.580 --> 24:03.180 From the very first version, the possibility of permanently tracking the user was considered. 24:03.340 --> 24:10.960 As you may see, permiso de ubicación means access location permission, and todo el tiempo means all the time. 24:11.580 --> 24:15.980 The application has far set of data that try to run a startup. 24:16.120 --> 24:22.000 For example, show far set of date dialogue and device startup or on boot. 24:23.680 --> 24:29.120 Also, this location tracker attempts to listen for boot events. 24:31.160 --> 24:34.800 The app tracks and asks for medical history. 24:34.900 --> 24:43.620 For example, cancer, diabetes, pregnancy, cardiac hepatic, renal, respiratory diseases, and all this data is stored abroad. 24:45.100 --> 24:50.920 Auto-evaluations are not stored client-side, but rather sent to the server. 24:53.340 --> 25:01.920 The user's location, along with his or her DNI, the national document ID number, are sent to a remote server. 25:01.920 --> 25:09.940 Again, this identification number is like the social security number of U.S. 25:09.940 --> 25:11.180 residents. 25:12.240 --> 25:21.220 If the user's evaluation returns that he or she is infected, tracking service is activated in the background. 25:25.810 --> 25:33.090 Also, by default, the app allows backup mode, which may send private medical information to Google. 25:35.250 --> 25:46.990 Now, it's unreliable for an application to diagnose a disease, but it's less reliable for an app to say, okay, you're not contagious, bro. 25:47.150 --> 25:50.950 We are during a pandemic, but hey, you're not contagious. 25:50.950 --> 25:52.250 You are safe to go. 25:53.290 --> 25:56.270 No sos contagioso means in English you're not contagious. 25:56.650 --> 25:59.910 So, the app has an option or feature for that. 26:00.430 --> 26:05.150 Now, remember the last time you Googled your symptoms of some sort, anything you had. 26:05.150 --> 26:13.190 I had a slight fever, my ankle hurts, I have a headache. 26:13.570 --> 26:19.490 What is the first result that Google answered you? 26:20.750 --> 26:27.090 You know, it's probably not something good at all or not accurate at all. 26:29.170 --> 26:39.230 Again, for example, in this snippet of code, it tries to determine if a user is not infected or not contagious. 26:39.430 --> 26:44.110 Again, you might have to see a professional, not an application. 26:46.490 --> 26:51.870 And this is not the first time we do something like this on the Argentine government. 26:52.010 --> 26:59.050 We were featured a lot of times on Programming Horror on Reddit for this kind of things. 26:59.230 --> 27:08.110 For example, this was the application that the Argentine government forces people to download when they land in Argentina to control coronavirus infections. 27:08.670 --> 27:18.930 As you can see, it is really long if it might give you a little sample of how we work here. 27:21.070 --> 27:29.850 Now, on the first version, the 1.0.2 that I have reversed, I found this string all across the code base, all across. 27:30.190 --> 27:31.890 So, who is Sergio C? 27:31.890 --> 27:33.990 We won't disclose its full name. 27:33.990 --> 27:35.070 Who is Sergio? 27:35.150 --> 27:38.230 A possible author, a project lead. 27:39.550 --> 27:45.970 After some Google foo, after some Google queries, I landed on his personal site. 27:46.390 --> 28:00.770 He has worked on a lot of government apps before, including transport, local government, national government, healthcare, and even now on the government's main application, which is MyArgentina. 28:03.340 --> 28:13.000 So, following the email he provided on his site, I found out that he was leaked a good couple of times. 28:13.000 --> 28:18.220 So, this might even rise the attacks on face for this kind of application. 28:19.680 --> 28:23.440 Now, what is everybody else doing? 28:23.440 --> 28:26.060 What are all our neighbors doing? 28:26.780 --> 28:33.060 All of them are dropping centralized solutions or centralized contact tracing. 28:33.620 --> 28:40.160 But we insist on keeping this model and pushing forward this model. 28:41.400 --> 28:50.060 And so, to close this chapter, this application was launched without publishing an audit, report, or penetration test. 28:50.060 --> 28:54.300 And it isn't even HIPAA compliant, not at all. 28:54.300 --> 29:02.100 This is especially serious knowing that it handles medical information and sends it to servers abroad. 29:02.360 --> 29:10.540 Now, ask yourself, please, what will happen if you develop an application of this poor quality for your employer? 29:12.870 --> 29:17.030 Let's jump to the conclusions and questions and answers. 29:18.130 --> 29:31.390 Although it may seem obvious for us, who are techies, geeks, hackers, or whatever, always inform yourself through professional, neutral, and verified sources. 29:31.870 --> 29:39.070 They are not easy to get, and again, it may seem obvious to us, but not to the rest of the world out there. 29:39.130 --> 29:49.350 Just because a term or a phrase is trending does not mean that it is real or automatically represents the thinking of the majority, right? 29:50.450 --> 29:59.250 Every day, somewhere, and at all times, there are groups of people and machines designed to install biased thoughts and debates in society. 29:59.810 --> 30:16.110 Surveillance and monitoring of citizens is not the answer, particularly in the hands of governments that have constantly committed systematic abuses against freedom of expression, especially in digital media, and obviously, especially in Argentina. 30:16.110 --> 30:19.110 You have seen some examples in this talk. 30:20.030 --> 30:27.090 Investing thousands in technology and applications after neglecting the health infrastructure is not the solution. 30:27.370 --> 30:31.810 An expensive app does not replace a doctor at all. 30:33.210 --> 30:44.190 You can get in touch with me via Telegram or GitHub, where these slides will be shared, and the video that I didn't have the time to play. 30:44.370 --> 30:48.330 And feel free to follow me on Twitter at Mauro Eldritch. 30:48.730 --> 30:53.330 As you may see, I have a few followers since I'm not into trolls, actually. 30:54.070 --> 31:00.850 I really hope you enjoyed this talk, and if you have any questions, I'm glad to help you and to answer. 31:01.150 --> 31:05.690 So please feel free to drop in and ask whatever you like. 31:06.010 --> 31:10.210 Thanks for watching this, and I hope to see you again next year.