Welcome everyone to another installment of the cybersecurity lecture series here at the School of Cybersecurity and Privacy at Georgia Tech. Today I am honored to introduce Anita Gohdes, but I am not going to do the introduction. I'm going to hand it over to fellow professor, my colleague Nadia here from Georgia Tech, SCP. So take it away. Thank you Brandon. I'm very happy to welcome. I need the goddess who is a dear friend of mine and collaborator. And to virtually to Georgia Tech. Hopefully she'll have visit our wonderful campus in school in the near future. But a need that is one of the few people who started working on the topic of cyber in the discipline of political science. Though she was one of the trail, trail blazers and done a wonderful research. Though Anita is a professor of International, a cybersecurity studies at the Hershey School in Berlin. Before Oberlin, she was at the University of Zurich. Prior to that, she was a postdoc at the Harvard Kennedy School and Liz International Security Program. I need. So I got her PhD from the University of mayhem in Germany. And her work has been published in various political science journal, tell political science journal. But one thing that might be relevant for our students is that show work has been published in nature human behavior as well. So her book project focuses on how states, how a particular autocratic regime used cyber as a tool of oppression. And today she will actually give us some very brief introduction into this approach is I'm very excited again to welcome her. And anita, you can start whenever you're ready. Thank you so much for the extremely kind introduction. I'm delighted to be joining you. I wish I could be there in person. Hopefully COVID permitting. I'm soon. And I'm especially happy that you're introduced me now. Yeah, because I consider you one of the trailblazers and low-skilled. And so I'm delighted to be joining you and your students. I'll be presenting at parts of my book project, specifically focusing on one aspect of what I call site, but it's domestic cyber controls, which is Internet shutdowns. But I'll be kind of hinting at broader questions throughout. Just to just to say up front, I hope that everything works with the slides. I had some trouble sharing my screen. So if you see some wonky this in-between, these just ignore it. It's because i'm, I'm sharing my slides separately, but it'll get started now. So the, the kind of starting point for this project was thinking about how it's at mass Internet adoption has really changed over the last couple of decades. So what you see here are average number of individuals using the Internet between 1990 and 2017. And if you can see, I'm on the side of the lines. I'm looking here at averages across different types of political institutions. And it's kind of unsurprising that already in the early 2000s we see this huge increase of Internet usage, specifically in liberal democracies. And for a long time and the kind of early 2000s, up until the mid 2000s, some of the less democratic countries were actually a bit skeptical when it came to rolling out internet access for everyone. Now some of that of course, has to do with economic development and infrastructure. But we know that countries such as Saudi Arabia were initially a little bit hesitant as to whether they should introduce the Internet for everyone. Because of course it comes with questions of having to control content. If you don't want your citizens to be able to access all types of content. But if we then kind of look from the late 2000s to beginning 2010, and then really kind of taking off in 20141516, we see specifically that closed autocracies and, and also electoral democracies are really increasing their Internet, Internet access. And so the question kind of is, why do we see this increase among countries that might have something against the broader population having access to the Internet. One of the things that we've seen is that even though we've had this increase in internet accessibility, that gives people access to all types of information and allows them to collectively mobilize governments by enlarge to remain in control of Internet access. And so I really like this picture from the Hong Kong protests in 2014. Just as a kind of reminder is that protesters need infrastructure. I mean, here we see kind of a visual representation of electricity that we can think the same also applies to, to the Internet. So work by people such as wrongdoer. It has shown that one of the reasons why autocratic regimes have been interested in adopting the Internet within their own countries is that online surveillance and censorship technology really has proliferated. And we've seen in particulate repressive countries have been very interested in expanding Internet access while making use of these types of technologies or what we might call domestic cyber controls. To the question that guides my research here is, how do these cyber controls actually end up informing date repression? Why am I interested in that? Well, if cyber controls can help maintain political authority in non-democratic countries, how does it actually into play with more traditional forms of political control? And one of the main ones that authoritarian states make use of is, is repression. And so that's kind of the research that's guiding me here. I speak a lot about cyber controls. That can mean all kinds of things. My, my definition here for cyber controls is that we can kind of situate this technology along, along a dimension where on the one side we have online surveillance. So that would be intersection technology monitoring of online information communication. You know, I'm collecting metadata and so on and so forth. The other end we have technology that's geared toward censorship. So restricting access, blocking access, ruffling Internet speed, filtering content, blocking content all the way to shutting down the internet. And so sometimes there's a bit of an overlap and what the technology can do. But in general, we can kind of think about this as a dimension from the valence to censorship. So what I'll be focusing on in this talk specifically is the logic of online censorship. And in particular the logic of Internet shutdowns. To the main argument I make is that if we want to understand the role of online surveillance and censorship, we have to look at how it is transformed traditional forms of gathering and controlling information. So ultimately the argument I'm making here as an argument about it. These are new tools for information control and we need to see how they actually differ from traditional forms of information control. And here specifically, I'm interested in how these information control methods help raise the cost of collective mobilization. And so we have to kind of think about the role of information and repression more broadly. Now, I'll briefly talk about the theory and then give you some kind of a snapshot into various types of evidence that shows us that this relationship between internet outages and state repression is, is quite an interesting and dynamic one. And so I'll start out with a global analysis of internet outages and state repression, showing some global evidence from 2017 to 2020. And then I'll zoom in on the Syrian case where we'll look at how nationwide Internet shutdowns are actually correlated with mass repression in the context of a civil war. And then we'll look at the Iranian case specifically at the Internet shutdowns in November 2019 and see how shutdowns are integrated into strategies of oppression in the context of mass protests. I'll talk a little bit about the conclusions at the end. So if we think about what state repression is, repression. And this is here drawing on work by people like Christian Davenport, is, can be understood as physical sanctions by state actors aimed at deterring individual collective mobilization against the prevailing political and social order. And so usually the way that, that is done is by raising the cost of collective mobilization or even individual mobilization. And broadly, we can distinguish between two types of repression. We can distinguish between targeted repression where the receivers of the violent coercion. So the targets are selected based on specific individual or collective characteristics. So you're a member of an ethnic group, you're a member of a political party. Those would be the types of characteristics that would make you a target for targeted repression. I'm, I'm targeted for pressure. Indiscriminate repression occurs when the violence does not discriminate between individuals or groups that are considered a threat. So in some situation, states might quote unquote, randomly attack certain parts of the population. For example, anyone who's out on the streets during a protest might be seen as someone who is challenging the state. And so this distinction helps us to think about how information plays into the dynamics of state repression. So let's think a little bit about the assumptions of that we have about the role of information and repression. The way that we can do this is by putting ourselves in the shoes of an autocratic leader. And so it was an autocratic leader. I have limited resources and ideally I want to use repression in the most effective way possible. And so effective repression tends to require information because ideally what I want to do is only target those people or individuals they pose a threat while sparing others. So I want to collect as much information about potential threats as possible. At the same time, autocrats generally assume that a certain level of censorship is necessary. The different, you know, different autocratic states use different levels of censorship in some of the media is more free than in others. But in general, there's an understanding of full Media Freedom. An old truck receive would lead to individuals and groups kind of having access to dangerous ideas, criticism of the regime. And this might insight collective mobilization against the regime. And so they're kind of two ways in which information functions here on the one hand, obtaining information for the sake of the regime and on the other hand, censoring information that the, the broader population has access to. And so you can really kind of think about how these, these modern tools of information control might play into these classic Dynamics or dilemmas. And the one we're going to talk through a bit more detail now is the difference between traditional forms of censorship versus online forms of censorship. So if we think about traditional censorship, media censorship. And this is something that still happens very much. Regime might shut down a media house. It might start regulating the content they're allowed to post that might start restocking the media. So putting other people as TV speakers, for example, or really changing the entire staff of a media outlet. It might flex them with new policies or laws or regulations that basically makes it difficult to work. And the thing with traditional media censorship is that it's very, it takes a lot of work to put, put in place and it's slow to reverse. So if you look at traditional measures of media censorship, they don't change a lot over time. Once you've reached often entire media outlet, it takes quite a long time to reverse that. So that's, that's a characteristic of traditional media censorship. Another thing about traditional media censorship is that the type of methods you use to target actual media outlets as an institution are oftentimes different to the ones you use to keep individuals or for example, groups from, from, from speaking their mind or from exchanging information and distributing information. Um, so while the one might be targeted by shutting them down, individual group censorship would work by repressing people, intimidating them, throwing them in jail and so forth. And so you actually need two different types of methods. When you, when we're talking about kind of offline censorship. A very resource intensive. And there's two of us. When we think about online censorship, things are quite different. So if you think about domain levels are, for example, shutting down the New York Times am, or what I call top-level censorship. So shutting down all access to the Internet. That can actually be quite straightforward for, for autocrats, right? Oftentimes they have very good connections to Internet service providers. Or they actually are partly controlling state or have state-owned internet service providers. And so that actually makes it quite easy for them to coordinate, a relatively cheap for them to coordinate these types of censorship activities. And another feature is that the same tools that can be used to, for example, shutdown. The New York Times or a media outlet can also be used to shut down discussions on, for example, a Messenger app. And so it's the same technical procedure and it's one that can be reversed relatively quickly and so wrong D burden and colleagues at the Citizen Lab oftentimes talk about this being just-in-time information restriction. So just before an election, you might want to do that where generally you actually are interested in having free discussion online, but then you can kind of shut, shut it down just in time. So that gives you as an autocrat, much more variability in terms of giving people access, removing it. I'm taking them by surprise and so on and so forth. And so that already has kind of been advantage of online censorship. But it also comes with a drawback. The drawback is that when we compare the effect of censorship on the ability to collect information as a, as an autocrat. There is no direct link in traditional censorship. So we can assume traditional information gather, gathering happens through security services. So I'm based in Berlin, goes home for my home of the Shaw, the one of the most well-known secret services engaged in information acquisition. And they're, the, they're activities of kind of in-person networks of surveillance didn't have a lot to do with with, with the type of media censorship that we saw online. That's very different. We see a direct relationship. If you shut down people's access to mess drops your censoring them, but you're also taking away your ability to collect information that people are exchanging online. And so it introduces a trade off that wasn't quite as strong. In the traditional setting. So we can kind of summarize it as the following way. You might want to prioritize online surveillance. And this gives you actual intelligence. It allows you to use more targeted violence. But the prerequisite is that you allow citizens to have a certain level of its accessibility. Whereas censorship can help restrict communication that can help keep at bay those dangerous thoughts, it can help hinder collective organization. If you think about specifically messaging apps. It can also help deplete opposition capabilities when, when the opposition is making use of the Internet for military purposes. But at the same time limit to access to, to intelligence. So this is kind of the dilemma that, that autocrat space here. So the implication I'll be focusing on in my talk today is that all else equal? We should expect that controls to Internet access though shutting down pot. So all of the Internet should be associated with an increase in indiscriminate repression. We should see that states are more likely to use mass violence or mass repression in the context of limiting access to the Internet. And we'll look at that in these three scenarios that I talked about before. In the book, I talk about the global analysis that I just talked about. Then I go a little bit more into detail of the development of this kind of digital control policy in Iran. And today I'll be focusing specifically on the 2019 Internet shutdowns. And then I take a deep dive into the Syrian case. I'm looking both at nationwide shutdowns and then looking at at the regional level instead accessibility in the use of parliament repression. And in the talk today, I'll be focusing on, as I said before, the global analysis, Iran and then just the nationwide shutdowns in Syria. So coming to the global analysis of instead outages, there's been a lot of recent interest in understanding the logic of internet outages or Internet shutdowns. But we have a couple of challenges would actually wanting to study this. And so most of our research, and that includes some of my formal work as well as focused on individual countries. And oftentimes the countries we're focusing on unknown to be repressive. And so if we're only looking at the effect of shutdowns and country that a repressive were kind of overestimating the effect of repression. Another problem is that we oftentimes focus on media sources to code these shutdowns. Um, and so oftentimes when there is a shutdown in a country during an election or in the case of Myanmar during a military coup, there'll be a lot of focus on the country. People will be trying to understand what's going on there. And so if there's a shutdown, everyone knows because journalists can't get access to that today, regular souls. And so if we're focusing on repressive countries and specifically on repressive episodes were possibly overestimating the correlation between shutdowns and contentious politics. Does we kind of selecting those cases out? And so what I want you to do here was look more broadly if we compare all countries and we use a network-based measure, bool For int and accessibility for shutdowns. Do we still find repressive patterns? I'm a kind of a general global level. So the, the, the measure I'm using here is based on data collected by the Internet outage action and analysis project, which is based at the University of San Diego. And they've been doing absolutely fantastic work. I'm collecting and developing their own net network measurements that help us to get a much better sense of the pervasiveness and the frequency of internet outages and beyond these media reports. And so what they do is they have their own network measure, which is what's called here active probing. So what it does is it probes the fraction of the entire IP space to see where they can basically ping service all across the world. And then they have a measure that's based on measuring internet background radiation, which they term a measure of a dark net activity. And then the third measure is a mixture of monitors of the, of, of the PGP network basically of global Internet routing. And so these three measures are different ways of trying to see where the Internet is accessible across the world. And they do this basically, ADA, at the minute level across the entire world. So what they do is they define outage events by comparing for each data source current values of accessibility with historical values. And if they see a significant drop between previous values and current values, then they code that as an outage. And so we have very, very fine-grained information on accessibility basically of service around the world. Now, the fantastic thing is that they actually provide an API. So if you're interested in doing work on this and you don't know the iota project, I definitely recommend going to that website and looking at it, it's fantastic, fantastic data. So what I am and you can basically get from the API information on all these outages. So I queried their API for all outages, for all three meshes. But every country basically all the data I could get my hands on. And they stopped their measurement and mid 2016 and I so I start collecting data and 2017 to 2020. And now one of the problems, of course, is that it's kind of dependent on how big the country is, to what extent we might have measurement error in there. And so I exclude countries with less than 0.5 million inhabitants, the kind of accounting for some of that measurement error. And then creates a very conservative measures of internet outages. Because of course, some of this might just be tech, technical blips or other types of outages. And so in order to reduce the number of these technical outages and be able to focus on the more long-term shutdowns. I only define a day as having had an Internet outage when there are at least two measures out of these three that indicate that there was an outage that was at least 24 hours long to some kind of network production for at least two measures that was at least 24 hours long. I aggregate that to the year because unfortunately, those of you who do social science research, a lot of the measures that we have for repression are at the yearly level. So I'm here aggregating to, um, to, to the yearly level. And I have two variables. The one is there was at least one outage day in a given year. And then they were at least 30 outage days in a given year to kind of see if there is, if there is different there, if we code this even more conservatively. So just to give you an idea what this looks like, this is an Internet outage and got born in January 2019. And you go to the website of iota, you'll basically be able to see this. You'll see in red here the period that they have identified as being an outage. In green here, you see the BGP measure. In blue, you see the active probing measure, and then in red you see the document measure. These are just some descriptive statistics for us to look at. If you remember, I started the talk with these, these, these categorizations of political institutions by country or political regimes. And this is based on Varieties of Democracy data, the regimes of the world, classification of countries where we have liberal democracies, electoral democracies, autocracies, and closed autocracies. And we see that across all of these cases, across the years, they were actually a lot of countries that had a lot of outages. But what's interesting here is that specifically the countries in the middle, so electoral autocracies and democracies are the ones where we see the biggest increase across this time period to as Internet access is kind of I'm increasing in these countries. We also see that there's a lot of kind of dynamic use of Internet outages. And we also see a big increase in these enclosed autocracies as well. That doesn't mean constant censorship, that means short-term forms of censorship. Of course, I want to look at this in a multivariate setting as well. So because I'm excluding smaller countries, I look at a 168 countries. And the way that I measure repression is looking at the political terror scale, which I'll show you in a moment, um, in case you're not familiar with it. I'm looking at Internet outages and then I have a set of control variables that we know from decades long research are really, really important predictors, full date repression. So the idea is, if we control for armed conflict, the size of the country, how wealthy it is, how democratic it is. And then we also account for yearly fixed effects. Do we still see that Internet outages are positively correlated with an increase in repression. So just for those of you who don't know the political terrorist scale, this is based on coding of the US State Department Human Rights Reports. And it's basically coded in five categories. The first category, secure rule of law, That's basically very, very little repression happening. And then it kind of increases until the level five is basically state terror. So with the entire country, we might talk about indiscriminate, indiscriminate violence and then the categories in between art and the nuances of that. I am, I'm a big proponent of these types of measures cross nationally because if we go any, if we disaggregate farther than this at the, the kind of country year level, we're using even more measurement error. So it can make sense to kind of use these more crude measures. Because ultimately, if we can detect a difference here, then they must be actually quite a big, quite a big change between the, between the different categories. What I'm showing you here is the marginal effect of a country year that has at least one day of an Internet outage compared to everything else being held equal, but there being no internet outages. And so what we see is that countries that have an Internet outage, or significantly less likely to be in Categories 1 and 2, significantly more likely to be in category 3 and 4. And then we basically see that for highly repressive countries, there's no real difference to be found. But basically what this tells us is that at the global comparative level, there is a correlation, even accounting for other factors that explain repression. That tells us that internet outages are actually correlated with high levels of state repression. This is kind of the big picture. And now we're going to zoom into two cases and look at those in more detail. The first one is looking at nationwide shutdowns and repression during serious civil war. And this is of course, not your average authoritarian kind of daily, day-to-day business where you might have some protests. But this is a situation where you already have a mobilized opposition that you're fighting again. So the assumption I'm making here is that mattress repression is part of a concerted military offensive. So we're looking here At seven nationwide shut downs that occurred between 2011 and 2014. And my argument here is that the main goal was to stifle the rebels military capabilities and isolate them from their local and foreign support base. And so we see that in the context of of mass repressive offensives, these Internet shutdowns are giving the government an additional advantage in their, in their military strategy. That I can show you three examples here. And I think what this also nicely shows is that just because there are national level shutdown, that doesn't mean that the effect is going to be across the entire country. Usually what's happening specifically in the Syrian civil war. That violence was oftentimes, I'm very localized at different times. And so the top the top row shows you during the the nationwide Internet shutdown in June 2011. Most of the increase in violence was focused on the governor of Hama. In July 2019 in 2012, most of it was focused on the area in Damascus and around Damascus. And then if we look at intense fighting that occurred in early May 2013, that coincided with this nationwide Internet shutdown while the government was trying to regain strategic control of certain territory in the South that included strategically important trade route to the cell. And so it shows us that even though the national level shut down mean that the strategies are nationwide, they can still be relatively localized. So to show you what this looks like, if we, if we kind of abstract this and look at it over a longer period of time, of course, we only have seven cases that we can look at here. So a lot of it might be to chance. So what I'm doing here is basically using a placebo treatment and saying what happens if we move the, the, the treatment by 30 days backwards and forward? Do we find the same effect? Maybe it's just kinda due to chance. And what this kind of shows us is that Internet, basically the day before an Internet shutdown, violence or documented islands peaks. And it stays very, very high during the shutdown. Then basically in the time before and after we don't find this correlation. So this gives us some confidence that we're not just talking about something that's cute into two chunks. And what I'm showing you here is documented violence. And you can of course think about picking, availability of information through technology being correlated with our ability to collect information on what's actually happening on the ground. And so one of the things I do in this project is estimate levels of underreporting to how much of the violence. Occurred with actually reported and how much of it wasn't repulsion. And so what you can see in this next graph is basically for every shut down in the top-left corner, you see it kind of aggregates is across all shutdowns. To what extent was violence underreported in the days leading up to the shutdown? During the shutdown and then after the shutdown. And what you can see that there's quite a lot of variation across the different shut down. So it's not that every time there's an Internet shutdown, we have an information vacuum. Nothing happens, right? But on the whole, we do see that there is a certain level of reporting that is missing during those shutdowns. But so does it happen before and after the shutdowns as well? So what this tells us is that the patterns were picking up are likely underestimating the violence. But they're good. They're giving us a clear sense of repression being most intense in that very short time period around shutdowns, but oftentimes happening a little bit before that already. And so you can think about this in the following way. If you were an autocratic state and you want to absolutely make sure that no one is finding out about what you're doing. You probably shut down the Internet first. And then you'd go and commit atrocities. That this is not what we're seeing in the Syrian case. The reason for that is that we are ready in this case talking about an active OMD internal conflict where the state might be less interested in saving face. That is a, that is different from the case I'm going to show you now, which is the case of Iran. Iran. We have a case of a nationwide shut down that occurred during mass protests. Very different situation. We're talking about an unarmed, relatively spontaneous protest movement. To, to tell you a little bit about this case. On November 14th, 2019, the government announced a 200 percent increase in the price of petrol. And this spot almost immediate outrage across the country. And the protests round this spread incredibly quickly across the entire country. So this was not just in Tehran, this was the protests that occurred in lots and lots and lots of different cities across Iran. And within a day, this, this is this movement. It really kind of already gained huge, huge tractions. And a lot of the diaspora were reporting on people using, for example, weight, the Waze app or Telegram app to coordinate where they would meet to, to, to produce traffic jams, to basically slow down traffic within cities and, and kind of coordinate their approach has tactics online. And some of the early evening of November 2016, so really just 1.5 days off to the protests started. The government shut down entire access to the worldwide web. And what you see here in the graph is basically again, just the data from, from the IoT project showing that basically there was, there was hardly any signal to be picked up from, from, from Iran during this time period. Now what's interesting about this case is that even though access to the World Wide Web was shut down, access to the national intranet. The Iranian Internet actually remained intact. And the government has basically since the mid 2000s, what quite actively I'm building this Intranet. You to have mobile businesses and e-commerce and mobile banking be, be functioning on the national level, intranet. And so when we think about the trade-off between economic trade-off of shutting down the internet. If you can leave on your intranet that provides access to some of the commercial activity within the country than the costs aren't quite as high to shutting down the worldwide web, specifically shutting down access to things such as Telegram. And so this kind of reduces the, the trade-off that, that autocrats have here. I'm a little bit, and that's why I think it's something we really need to pay a lot of attention to. Now, the, the strategy and specifically the outage was accompanied by mass repression by the security and paramilitary forces. So I worked on a project with, with Amnesty International to kind of brace the the ways in which the shutdown impacted both the human rights situation but also the reporting of human rights situation. So there's a lot of evidence of tear gas being used, live ammunition being used, and specifically this evidence supply of ammunition being aimed at protocells, vital organs, which suggests a shoot to kill strategy. So we're not just talking about regular protests policing, we're talking about a real violent repression of these, of these protest movements. Throughout this time, thousands of people were arrested and detained. And later there was evidence leaked that the Supreme Leader met with his direct security forces on the 17th and said. Do whatever you need to do to, to, to end this. We just need it to end it. Now. So we have a lot of evidence showing that mass repression was, was the tactic. But what's interesting here in the context of the of the shutdowns is that it was accompanied by a move towards silencing and intimidating victims and specifically the victims families. And so with a shutdown, what we see is security forces going to the families of, for example, committee members who'd been killed, saying, We're only going to release the body. If you if you promise us that the, um, the funerals going to be held very small, it's low profile. You don't post about this anyway, don't tell anyone. And we're really pressuring people to not speak out and draw attention to whatever happened in the family. Um, at the same time we see classic kind of date media report saying protests, riots as it's all foreign conspirators. And there the violet ones and all we're doing is kind of defending ourselves. And importantly, there's a lot of information I'm pointing towards people's mobile devices being confiscated during the protests so that even if they're recording things, they can't automatically stream them. The devices are being held so that they can't release the information later. In some of the the the international reporting said we actually have a time lag of about 24 hours in terms of being able to actually report on these event because it's so difficult for us to get contacted people within the country. Amnesty International and many, many other human rights organizations have tried to establish a death toll of what happens specifically in this time period and after the, the shutdown as well. And that death toll is still not fully establish. So and the numbers range from somewhere between a few 100 to a few thousand. I mean, it's not, it's not exactly clear. And I think what's interesting here as well as, so The amnesty around team did a lot of work on collecting video evidence from activists within the country, from, from their own networks, right? And so what you see here is at the hourly level video evidence collected by the amnesty team via local sources. So there's a ton of evidence coming in hourly right before the shutdown. And then it really kind of thins out to the strategy of creating information vacuum to a certain extent is working. In this case. What's interesting and this is just kind of a byproduct. But if you think about the reactions to this type of censorship, We have some information. For example, there's a, there's a great study by it. I'm Molly hobbits, muddy Robinson. Well hubs looking at reactions to the blocking of, of Instagram and China and check that it actually led people to look for to convention software. And we find similar results here and an Iran. So I looked at specifically the use of Tor. Tor is always been very popular in Iran and both to surf anonymously, but then also to be able to circumvent censorship. And what you see here is basically the number of concurrent tool uses throughout the, the shut down, of course, during the shutdown. Or they're basically none because there's no access to the web. But then once the Internet comes back on, actually more people using TOR because more people wanted to circumvent that. More people want to be able to surf anonymously and so on. But on the right-hand side you can see that there's really quite a significant difference in the amount of people trying to access a type of information. So in a way, if you think about the broad implications, this is, this is a strategy that has helped silence some of the purchase to a certain extent. It has helped keep some of the details more opaque. But at the same time, It's really motivated people to use this type of circumvention software. There. If we think about the conclusions more broadly. And what I really want to look at in this project was how these online tools of information control, I've transform traditional surveillance and censorship. I want you to look at how they're connected with United States digital infrastructures and how specifically author riparian states are thinking about these new forms of information role and how they were including them in their border repressive strategies. And basically nowadays we see very few protests happening, even in semi democratic countries where there's not some type of censorship occurring. And so this has become really a part of the standard toolbox of process management for, um, for many, for many states. We can also think about the implications for policy. When we see Internet shutdowns occurring, it's important for us to kind of think beyond, you know, the, the explanations that might be given by state. Sometimes they say, Oh, we're doing this to limit the, you know, the, the circulation of fake news. Or, you know, we doing this to limit islands. But oftentimes it's actually part of their own repressive strategy. I haven't talked too much today about surveillance software, but of course, a lot of this, this censorship or filtering software also is built specifically in the European Union and the US. So the export and the trade of this type of software has huge human rights implications. And lastly, of course, it has big implications for the resistance toolbox of the activists. To thinking about, you know, I'm, I'm empowering human rights defenders on how to use the convention software, how to revert back to traditional forms of communication in the context of information blackouts. Those are all I'm really, really important topics for us to think about. Okay. I thank you very much for your attention. That's also me and I'm looking forward to your questions. Wonderful. Thank you so much. We've got some time for questions. Any questions from the audience? I have one from the virtual audience actually that I can read out. Stowe. Would you consider the tactics of the Canadian government is taking against the for the freedom protesters in Ottawa as both surveillance and censorship. When it comes to tracking down what donations were made to the protests and subsequently firing people were cutting off access to private bank accounts. I have to admit I've been following the protests that haven't actually been following the, the information control aspect of these protests. That what you're describing in terms of kind of digging into people's history. It's definitely part of of trying to identify who these challenges are. So I would I would classify that as a type of surveillance and some events endeavor. Absolutely. But I haven't study that case a more but now I I definitely want to it's very interesting. Yeah. Thank you so much for that. Any other questions from the room, anything? Things like our virtual audience is more active today because I've got another one from the Internet. Someone else asked, what type of Internet censorship do you find most effective besides a total shutdown of the infrastructure? And how might AI play a role in censorship and surveillance? Though I think an increasingly popular tool if we think specifically about the concept of plausible deniability. So oftentimes, semi, semi democratic with semi autocratic states are interested in using repression, but kind of getting away with it without people knowing that they using it. And so I think in the context of plausibly dynein your involvement, the throttling of internet bandwidth is something that is very effective and has been used also by the Iranian government, for example. Because the problem is, for those of you who've ever been a big gatherings of protests, those the kind of accumulation of people in a certain area can already lead to a reduction of bandwidth that you have, right? And so it can be difficult to distinguish between intentional throttling and throttled speed as a consequence of of the large gathering. And so ruffling can be really effective in hindering people the ability to upload, for example, videos, which is oftentimes what you want to avoid, right? Video and picture material because those are the things that have a big effect on people's perceptions of conflict. That you can do that without people saying you shut down the internet. I think that's, that's a tactic that we're increasingly seeing. Luckily, computer scientists, I know that many of you are computer science have actually develop tools to, to, to try and distinguish between intentional throttling and throttling as a byproduct of of bandwidth reduction. So I think that's, that's important for us to be able to distinguish those things and detect them. Yeah. Thank you. Thank you very much. If you don't mind, I'm going to continue on a point you made. So I think it's very appropriate that this presentation was given at a place like SCP here at Georgia Tech because one of our driving missions does disorder combine the technical and then more broad reaching impacts of the cybersecurity questions. And I really appreciated how in your talk you through our credit. So the group at UCSD, computer scientists, they built these scan, scanning infrastructure. I am personally very close friends with the computer scientists, of course, at UCSD that designed those techniques. I'm curious from your perspective as someone who kind of takes these datasets and runs with them. What more could computer scientists and ends very technical cybersecurity researchers do to provide that data that you can then draw more wide-reaching research from. That. That's a wonderful question. I love that and thank you for all the work you're doing. And pistol, I have to say the team, it was incredibly helpful. Walking me through their API. Just providing an API that researchers can use, I think, is a huge, a huge service to the discipline. And and, you know, walking me through things such as, you know, you should probably excludes molar countries because of measurement error, those kinds of things. That was incredibly helpful for me. Just having someone who was supportive of me using that data was, was really, really great. I think I think the detection of intentional versus non-intentional behavior is something that's really, really important for us to being able to detect when such outages are the consequence of a technical failures or whether they're intentional work. So when early on when I started working on shutdowns, in theory, I was relying a lot on reports by Cloudflare and other other providers looking at how these shutdowns occurred and these experts say, oh, you know, this actually looks like coordinated behavior does look like there was some cable that was cut or something. So I think anything in the direction of can you tell us, is this human directed or is that something that's a byproduct of technical issues is incredibly helpful for us. The same for throttling, for example. And then just in general, I think data availability is fantastic. Crescent. Yeah, Wonderful. Thank you so much. We have one more from Lauren needed. This was really, really fascinating and I'm really looking forward to your book. This talk really throws a lot of cold water on this narrative of liberation technologies that we've heard so much about. And I'm curious if you see that debate is a simple either or, or maybe there's some temporality, right? It is really good for organizing dissidents and protests in the beginning. But then as states start to learn more about it than the advantaged, Swift's heavily toward the state. And a second and related question, I think this really bears on this larger question about fiber conflict. Is it a substitute for conflict in aggression in meet space, or is it a complement, right? So it seems to be that the information controls here really work best when they are accompanied by good, old-fashioned repression. Yes, absolutely. So I don't think it's I mean, I I'm luckily, I think we're, we're kind of moving away from the over-optimistic cyber Liberation Technology or the ova, pessimistic kind of repression technology angle. To think that there's a lot of nuance in there. I do think that there's a lot of learning that's happening specifically. I think the Iranian case shows us that. I think the Syrian case shows some learning and tons of strategies used by the regime and then certainly a lot of more kind of stable high-capacity authoritarian regimes that show us that there's a lot of learning that's happening there. But what I always like to remind my students is that we wouldn't see regimes going to this much trouble in trying to block access to information. Essentially, if they weren't something dangerous about information. And so I think as long as that is still the case, that is already alone. It's kind of an indication that the free communication is a danger to authoritarian regimes. And we consistently see that resistance groups are getting better at circumventing these things. Building more stable structures, being more aware of specifically surveillance. And so I think, I think it's it's not a baffle that's one by one side. That said, I am, I am actually far more worried. And the long-term about the surveillance technology aspect and the the, the, the censorship aspect. Because I think it's much more inside us and all the, just the capabilities that states develop with this. And then I'm definitely a, a proponent of the complimentary view. I don't think it any way that these things are replacing traditional repression. If anything, they're being kind of included in the broader toolbox of process management or dissonant management. Thank you. I have a question that talks a little bit about policy applications and you work with Human Rights Watch, Amnesty International. You've been involved with them since 2009, I believe, basically over decades. So I wanted to know more how you work with them, informs your own research, but also how your research informs their decision, right? How do they act on the ground? To give a little bit more practicality, I'll file or sort of research explaining practicality. Tell the students. Thank you. Yeah, I think it's been really interesting to work with specifically advocacy based human rights organization. Sorry, I shouldn't say that because It's human rights question specifically amnesty. They both do a lot of, and they're doing increasingly more research in this area. And I think a lot of the research we have currently about surveillance technologies actually coming from these organizations. I said this in lab and so on. So I think, I think they were, they were doing huge amounts of research. I always feel like the perspective I can bring in as a social scientist is to not believe everything you see in your data and actually critically kind of think about what's missing. So, so as an example, when we started working on this on this idea of trying to understand what happened in November 2019 and Iran. Um, my perspective was we really contrast the data. We have to see the data as a function of what, of what was observable. And so I think that that was kind of the perspective I could bring into the organization. They're much better at the verification process. They have, you know, all the, the contextual knowledge and all their sources and everything. And what we as social scientists can bring in is think about what's missing and how this is a function of information inaccessibility, right? And so I think that, that's something we can really contribute. Plus of course, I'm kind of quantitative skills specifically and kind of analyzing the data and looking at looking at breaks within, within, within the timelines and stuff like that. And then on the other side of working with the Human Rights Data Analysis Group, which is, which has a lot of them are statisticians and computer scientists. Also bringing in kind of this idea of how can human behavior tell us about what's missing in this data. But I think that's really what social science can, can, can contribute to the discussion. Yeah, thank you so much. I have another question from the online audience. So someone wants to know a bit more about the availability of these circumvention tools like chore, right? So is there specific education that people are seeing or is this sort of like a grassroots thing in these countries that need technology like to, or how does that sort of circulate in these communities? So I think that really depends on the country we're talking about. And it also depends on the level of just digital literacy, when might say. And so I think, I think Iran is a case where many people just use various forms of VPNs. A proxy is or tour every day. Because so many people make use of specifically Telegram. And, and the government basically does the most to kind of, I'm reduce peoples that tries to reduce people's motivation to use that, that app. In other countries, it's been a learning process, right? So I think just as the Syrian regime early on in the conflict didn't really know what to do with social media data. Initially, many people didn't use anonymizing software or Elias's online or, you know, we're, we're careful about those things. And so I think whereas oftentimes kind of full-time activists are used to using this type of technology. It takes longer for football to populations to kind of get used to using that information. But when, and that's one of the things, right? When the government limits access to a, to a website or a service that you used to talk to your family. You are very much blurred the circumvention tool, right? And so that's kind of the crux of, of, of the trade-off that the authoritarian has. Now if I shut down access to Netflix, people are probably going to do quite a lot to get access to that or Skype, right? But if I shut down access just to some obscure use website, who knows if they gotta go to the trouble of downloading a new VPN every, every week. And so I think that's kind of where we see it becoming difficult. Wonderful. Thank you so much. Excellent. Questions, answers. Thank you so much. I'd love to thank you again for visiting us here at Georgia Tech. That's the end of today's time, unfortunately. So let's thank our speaker for sure. Thank you so much for having me. It was a real treat. I'm sorry again about the slides. I hope you were able to to follow everything worked out perfectly. Thank you so much, anita, you have a good evening. Thanks, everyone.