 Welcome, again, to everyone in the audience. Welcome also to everyone watching via livestream from around the world. We have the pleasure to welcome Christoph Engelmann now. Christoph is an author, a teacher, and a researcher. He researches at the Institute for Advanced Studies on Media Culture of Computer Simulation. That is at the Leuphana University in Lüneburg. In his talk today, he will focus on the enabling dimensions of drone warfare. That's a little spin by especially looking at data gathering and the mapping of social graphs. We will have a chance for a QA in the end. So please take notes. If you have questions in the end, you can ask them. We have plenty of time, but for now, with further ado, please help me welcome Christoph Engelmann to his talk on graphs, drones, and phones, the role of social graphs for drones in the war on terror. Christoph Engelmann. Thank you. Thank you. It's exciting to be back. It's my third talk here. And as most of you know, I've previously talked about history and present day developments of identity media authentication technology. My last talk I gave in a pretty hungover state, which you will see if you watch the video. I slept better this night and had some coffee. Need more? Anyhow, what I'm going to do today is pretty much ask a very similar question than the questions I have asked before in this context here is how do you identify people? But here, in the context of drone warfare, because whenever the debate came to drones and surveillance in the past decade, I always wondered, how do they actually identify the individuals they see in the video feeds? Because that's an important part of the whole voluntary discriminate individuals and decide which ones you want to kill or capture and which ones not. And many may argue that drones kill quite indiscriminately. And that is definitely true. But at the same time, one can see, especially since the publication of the drone papers by the intercept in October, that it has a fairly elaborate process, a bureaucratic and administrative process, being put in place, at least in the American context, the process that is meant to positively identify the individuals you're tracking and eventually taking out. Many, many people involved, many, many hours of surveillance. There's this idea of an unblinking eye of looking at people for weeks and weeks and weeks and start the whole observation cycle anew once the eye blinks, once you lose visual tracking. But of course, this visual tracking somehow is suspicious. First of all, the quality might not be too good. Secondly, I couldn't fathom that they use biometric technology in the context of this video feeds because that would be too tall of an order. So there must have been something else. And that is what I'm trying. Part of this puzzle, I think, I can offer some answers to. There might be more that we don't know about. As with all of this research, it's difficult to do because there's little publicly available. You have to puzzle together from the Snowden files, from other leaks, and also from doctrinal papers or like BAMA and PhD works that people in the military do, which is an interesting resource. If you start to look at those, you will find tons and tons of interesting papers tackling exactly these problems. So One Little Clue came in 2008 or 2009 in this Wired article about tagging technology that's used in the war on terror. Basically, it's kind of RFID or ID tags or technologies that work in some way in the optical spectrum by tagging people with paint or any kind of other medium that then will show up in certain spectrum. Oh, the cat is gone. That's a pity. But that, again, seems to be fairly difficult to do because you need access to the individuals or things they own, equipment they have in order to do that. And also the question is how discriminate can this get. So there must be, as I said, something else. And in order to further my argument that I want to make today, which to outline that once more, focus less on the drones than on the infrastructure, I have to take you on a little detour through history more precisely the history of graph theory. And those of you who are trained in computer science or maybe some sociologist probably know way more than I do about graph theory. And I'm happy to learn from you in the Q&A or afterwards. The foundations of graph theory, this is Pregel and this is the River Pregel, were laid in the 18th century by the Swiss mathematician Leonhard Euler. And he was employed at the Academy of Sciences in St. Petersburg, which had posed the so-called Königsberger Brückenproblem. And the question was, can one cross the seven bridges of Königsberg crossing the River Pregel here and never cross the same bridge twice? And Euler's solution was basically to abstract this geographical ensemble into a set of points in their connections, which can be depicted like this or depicted like that. And both graphs, these are graphs, isomorphic, they're the same, even though they look different, they show the same relationship, the same topology. OK, and what we are seeing is four points called vertices or knots in graph patterns as well as seven connections, which represent the bridges in this problem, which are called edges. So the knots or vertices are the land masses and the edges are the bridges. And Euler could show that all four land masses in Königsberg would have an uneven number of bridges and that in such a setup, it is impossible to find a way that crosses the bridges only once. And graphists are typically notated like this. The graph is a vertex and an edge. And as I said, the number of edges between two vertices are usually called weights, and the more edges you have, the more weights you have in a given graph. And despite this graphical representation, I just showed here, it is important to remember that this is less a visual tool. It is also a visual tool, but less so, but for most a mathematical description where you can do calculations on. And graphs have found applications in a wide range of fields from chemistry where the relations between atoms and molecules can be represented as such in linguistics or in neuroscience where the connections between neurons are described by graphs. And some of you might have seen this presentation, it's actually by the NSA. They have two slides sets publicly available. They have a nice little journal for public consumption on their website. And in the 2014 volume on Big Data, they basically explained their interest in graphs. And this slide is taken from there. So this is what in neuroscience parlance you would call a connectome. It's the individual graph of the connections between the synapses of a given brain, which is a pretty big data problem, as you can see. And last but not least, of course, you have graphs and computer science where they provide one of the most important data structures used for file systems. It's the Linux file system, which is a tree, which is a special form of graphs. For dependency matters, control for representations, compilers, and so forth. But there's another important branch of science where graph theory has made an impact, in which moreover is important to the war on terror discussion and the argument I want to make here. And this is sociology. And let me just give you very short insight about the weird way this debate has taken in sociology. So they didn't enter sociology straightforward. As usual with new paradigms, they came basically from the sidelines. In this case, in the guise of Jacob Moreno, the flamboyant figure, most widely known for the innovation of psychodrama, which is a sort of improvised theater. We had a workshop on that or a talk on that at the conference here. And who also was an avid social reformer, engaged with convicts, with orphans, single mothers, and so forth, in the 1920s, 30s, and 40s of the last century. And alongside using psychodrama to educate and empower his clients, Moreno also developed a technique which he called psychological geography, or sociometry. And basically Moreno depicted individuals as points and their relations as arrow lines. And here you can see one of his psychological geographies from his book, Who Shelves Arrived, written in 1934. And that's a school class in their first year. The boys are on the left and the girls are on the right. And as you can see, there's quite a lot of connections going back and forth between the individuals in this class and between the genders. And basically Moreno mapped out these connections using, you know, questioning the people who do you talk to or how often do you spend time with such and such and so forth. And after two years, the same psychological geography looks like this. Boys and girls are basically strictly separated except for two individuals which basically cross the gender line here. And this already pretty much looks like a graph by Euler, but it needed a couple of trained mathematicians to make that connection. And cutting a longer story short, if you wanna read up on this, there's a nice book available for free online by Lyndon Friedman on social network theory where you can find basically the history of these developments. So this mathematician was Axel Bervelas who worked together with Kurt Levine, another famous psychologist of the first half of the 20th century at the MIT in the Research Center for Group Dynamics. And together with two PhD students, Richard Lays and Albert Perry, they provided the first formal definition or proof of such what we now would call social graphs. And more of a, invented more or less a class of measures which we now call centrality measures where you can mathematically measure the degree of connectedness and the importance of a given a vertex in a graph. So the context they did this research in was basically companies which had an interest in knowing the difference between the formal and informal hierarchies within their institutions. So this is at the height of fortism of Tayloristic work organization methods where you have very, very strict hierarchies, the boss on the top and the worker on the lowest level and many middlemen in there. And what the research showed is that despite this formal organization, you have very important informal networks. And early social network research was concerned with the relation between those two networks and with finding out who is actually central in the information flow in these organizations. Another important field of research in this context was pharmaceutical companies who wanted to know how to push their medications into the market. And for this they needed to know which doctors influence opinion. So they mapped out basically the informal networks between doctors in order to find out which doctors do you need to give basically or do you need to target for your advertisement. This is in the 50s and 60s. And this is the context where centrality measures first developed. They then move over into bibliometric research, basically finding out which academic sites whom, which is another example for more or less informal networks. So, let me skip a little bit here. Okay, so far for the historical outline, which served to highlight the context of the emergence of centrality measures from organizational psychology and sociology. It is the study of oral and written communication that is important in groups where these centrality measures emerged and where centrality became the index for the stability or instability of an organization. The tools of means are what we today call social networks. So many mathematical ideas on concepts emerged in these discourses and thus way before computers came into the pictures. So jumping forward to today, or basically, let's take that back, jumping forward to 2000, so 15 years back, we have seen the rise of a graph industrial complex. First of all, of course, Google, proved that graphs and graph centrality measures and PageRank is basically a graph centrality measure that came out of Leo Katz's centrality measure idea in which he invented in the early 50s, also in the context I just described. So Google proved that these measures are instrumental in managing the seemingly unmanageable dynamics of the internet and showed that you could graph out billions of websites and their vertices and their backlinks as edges and that you can provide a tool for navigation in this context. And since 2005, we have seen a number of new actors that emerged, which managed not only to capture the relations of websites but of individuals and their actions and graphs. Facebook, LinkedIn, Friendster, Instagram, Tumblr and so forth all rely on graphing out their respective populations. And in 2012, the consultancy gardener offered the following take on this development. They basically say there are five essential strategic graphs in today's economy. This is the paper which is paid content if somebody can provide me with the full paper. I'd be thankful, I only have parts of it. And so this is from this report. So the five strategic graphs, the social graph, internet graph, consumption graph, entrance graph, mobile graph. Search is probably intent or interest. That's a little bit unclear, right? But that's basically what Google owns. The consumption graph more or less is owned currently by Amazon. The mobile graph is split across a couple of telcos and the social graph, of course, is more or less owned by Facebook and Twitter, at least in the context of Western industrialized nations. So now most of you will probably make the connection. Is it just the economy that understood the important or the strategic relevance of graphs? And certainly not. So after the Snowden revelations, the German Bundestag, so the parliament, implemented an investigative committee to study the extent of NSA and Five Eyes Surveillance, especially in Germany. And part of the proceedings was the invitation of the NSA whistleblower William Binney, who also has been here at the Kaus Communication Congress I think two years ago. And asked by the German members of parliament what they actually did at the NSA, Binney answered the following. We built a relationship in what we call the graph, a social network of the world. So the transcripts of this sitting of the committee are available via WikiLeaks. It's 187 pages paper and it's really interesting to read because Binney basically explains back and forth how they graphed out the world and how important that was. He mentions graphs and social graphs, I think 15 times or 16 times. He never once gets asked actually by any member of the committee what that actually is. They either know and have a quite good understanding of this concept or they just ignore it. What they're interested in, if you read this paper, it's interesting to see, they ask a lot of questions about career pathways within the NSA on the one hand, so how do you become somebody like him, a technical director at the NSA? How much money he makes? He says 20% less than a U.S. senator, that's the highest pay grade in the NSA. And then of course they're very interested in how far Germany was targeted in this context. And the other interesting tidbit is that the NSA, at least he says, gave the Bundesnachrischendienst, the German equivalent, the source code to this project in the early 2000s. So, Bini's project was called ThinThread, which basically, we don't know exactly what it is, but basically seems to be a huge graph database. The difference to what came later and why Bini became an NSA whistleblower was that it basically did two things. First of all, it wasn't a full take approach like the NSA and Five Eyes do now, so they didn't take all of the internet traffic and secondly, it would encrypt the data of individuals with U.S. citizenship, right? If one is to believe Drake and Bini, Thomas Drake's the other whistleblower, the successor to ThinThread was a project called Trailblazer, which was implemented by private contractors and basically did away with these two restrictions that ThinThread had. So, that's their basic problem, right? I mean, it's astonishing how strongly they oppose basically Trailblazer, whereas ThinThread already, if you think about it, was basically graphing all the world and that's what he says, right? So, there's another interesting element in here and this is the timing of this. Bini mentions in the NSA and Tasung's report that they started in the early 90s with this project and had a working prototype in the second half of the 90s. To anyone familiar with the infrastructure and necessities of graph processing and the developments in the commercial field in the past decade, so what these companies I just showed from the Gardner report basically did. So, key value stores, map reduce and other means of distributed computing, you will agree that it's quite astonishing that they're already in the second half of the 90s where basically able to do this. So, the important part here is that it's not only businesses, but also governments that have understood the value of graphs and acquired the means of generating and exploiting them and I believe what has happened in the past 15 years, basically between 2000 and today, is an ongoing race to graph the world and not only a graph industrial complex, but the military academic graph industrial complex has basically emerged. On the commercial side, we have what Bruce Sterling calls the stacks. I just talked about them, Facebook, Google, Amazon and so forth and if you look beyond basically the Western world, you will see that China has Q-Zone, Tencent, Renren, Baidu, Sina, Weibo and the Russians have VK contactor, a more mere as Facebook equivalence and basically these are all invitations for you to become graphed, right? VK contactor, you get MP3s for free and they're happy to have you as a note in their system. On the government side, we have the NSEM-5Is which are in the business of graphing the world and the papers from the Snowden archives, if you look at them so far, only provide indirect material because they concentrate more on the means of access and collection of data. So Prism, Buran, Fairview and all these acronyms basically describe how to get the data and we know relatively little, at least on the basis of those papers, on the aggregation and analytical tools. William Binney, to come back to him last year in 2014 explained ThinThread and the Snowden files to select audience at a dinner in Washington and pointed out that Stellarwind, Mainway and Marina are the graphing tools used for discovery and development of targets by the NSA and this again are slides which you will find on the net. It's really interesting to read that through because it's basically a Q and A where Binney explains how this is supposed to work. We don't really know if he actually tells the truth but what he says is basically you graph as much as you can and then you use centrality measures in order to create what he calls zones of suspect is in the green bubble here and these individuals are the ones which you follow up further on and profile out. But what do you do once you have something like these zone of suspects, a graph of a terror network if you wanna call it that? And this is a question of much debate in both doctrinal literature as well in the academic research I mentioned in the beginning so the military academic research around so-called dark networks so this would be a dark network in the bubble here, the bad guys. And the striking example of the thinking so there's tons of papers I just pulled out too to show you a little bit how they conceptualize that. A striking example is the US counter-insurgency field manual published in 2006 too much public from far, I don't know how many Americans are here, maybe you remember because even the New York Times reported about this. The US military, I mean this is in the depth of Iraq and Afghanistan wars, they had no counter-insurgency doctrine. The last doctrine was written in the 70s in the context of the Vietnam War and then after the way that war when nobody wanted to do coin as they call it, counter-insurgency. But three or four years into these wars they figured out we probably needed doctrine. So they had this guy, right one, David Petraeus, who later became the CIA director until he was basically removed from that position in 2012. This is another interesting book which you really should read even though it's an ology and Fred Kaplan is not a good author, but he's one of those military experts. But if only half of the stories about the Iraq War, he tells are true, it's pretty amazing what happened in the political system in the US in that time. Anyhow, so Petraeus writes this counter-insurgency manual and of course there's an appendix on social network analysis and exploitation in there which is really, really short, just a couple of pictures. And here's an example how they envisioned this. So basically what you see is a graph of some kind of insurgent terrorist or whatever you call it, subgroup, which they map out. And now the important doctrinal notion is shaping. So what you wanna do in a counter-insurgency or a war on terror context is shaping the graph of the enemy. So you have the network on the left side, then you implement some measure like coordinate search, then you grab out a couple of people to start food distribution and during this time basically map out the communications, graph out the insurgents. And the idea is to shape the graph in a way that it reveals information about who is important in this context. If you can use of course the centrality measures I mentioned in the beginning. So again, this is shaping and there's much debate about how to do this actually. And another interesting example for what I just described is this paper that came out of West Point Network Science Center. They are shaping operations to attack robust terror networks and they use a publicly available data set, the so-called Tanzania data set, which is a graph of Al Qaeda 1998 in the context of the bombing of the US Embassy. And what they offer is basically an algorithm that promises to automate the decisions which nodes to attack in order to shape your graph. And the idea is to render the graph more fragile by generating more centrality, by making it less distributed and more central. So basically the algorithm tells you which nodes do you need to attack in order to have a star shape network in the end and then know who is actually the most important person in here. Because I mean the quality is bad here and it's really tiny in the paper. But basically in A they don't know who Bin Laden is and they wanna know who that is and this algorithm basically offers attack this node, this node, and this node and eventually the communications will lead to Bin Laden which is actually this individuals down here. So they call this fragility which is basically the, not really the inversion but tied to centrality and which seeks to find a set of nodes whose removal would maximize the network's wide centrality. And we also include the problem of no strike list. So basically the algorithm will take those people out which are not supposed to attack. And this is because real world target of insertion networks often includes restrictions against certain individuals and so forth. They also even prove that this is an NP complete problem. So read this paper, it's really interesting and there's many, many more, I think this is one of the more important ones. So this is basically an algorithm that helps decision making in shaping graphs, right? So, but how do you actually get those graphs when people are not online all the time like we are and have those little cell phones in them and now on Facebook and so forth? And that was the problem basically that you had in a country like Afghanistan and what the US military did was basically to hire anthropologists. They sent anthropologists in the field and asked people those questions here. What are, so the people in the villages and so forth. What five people here have you known the longest, et cetera? And you enter that in a tool like this where you basically generate a graph of the given population in your area and hence can shape it. This is the map, HD map human terrain. They called this the human terrain system which was closed down a year ago. It ran from 2006 to 2014, very controversial. The American Anthropological Association actually came out to oppose the system and protested strongly against it. There's an interesting movie about it which I recommend to watch. It's actually not clear, I'm not sure if they actually used the software. This is just taken from a handbook of the map human terrain system. But what they definitely used and what you will find in many, many papers is IBM's analyst notebook which is basically a software tool which offers you to map our social graphs and you get the centrality measures and can use that in order to track and profile individuals. And of course what they use is Palantir which I think owns the market for graphing technologies in this context. Anybody here who actually has worked on a Palantir workstation? I'm still waiting to find somebody. I strongly encourage you to watch the YouTube video on counter-terrorism that Palantir has up. It's really interesting. So Palantir is basically a Peter Thiel-founded backed startup founded in 2006 which has seen a meteoric rise and which started out with a software product allowing graphing for government entities and now has branched out in all kinds of fields, mostly fraud detection in banks but they also do philanthropic engineering. So if you want to map out your clients as a philanthropist, Palantir has a solution for you. Okay, so this is the analytical level but of course, what about the drones? And what I think we have to think about drones as is less tools of visual collection but more tools of collection, of communication metadata, basically cell phone data. And if you look in all those presentations which float around on the web, you will stumble upon all those pots that the drones carry. Of course, they always have a camera and they have a visual tracking tool on board but they also have the communications and networking. This is the Argos pot which is a wide area surveillance system. They have the communications and networking spot which basically siphons in all the communication data in a given area and that data gets aggregated and graphed out. So I think we have to think about drones as crawlers just as the Google bot is a crawler that indexes web pages and their links. What drones basically do is they crawl an area and index all the communication that is going on allowing you to build up a graph of that. Of course, there's also the other side. They not only collect, they also target and notes of the graphs they have created. And this is another strand of the doctrinal debate that has been going on in the past 10 years because the initial armament of drones were the Hellfire missiles which were anti-tank missiles developed in the 80s which have quite a strong warhead where you have the problem of collateral damage. And what has basically happened since 2005 is a race to find ammunitions which allow you to target individuals. So that's called discrete effects, right? This is, so personal targets, the fragment is personal targets plus less than lethal mode so you, I don't know what less than lethal means if something like that hits you. And I mean, this is just one example. There are many competing systems. This is called Viper Strike and they all come with this nice little circles which tell you the blast areas of different ammunitions and basically, as you can see, it's relatively small compared to a Hellfire missile or other ammunitions. So this is basically what shaping can be. There are two ways to shape a graph, kinetic and non-kinetic. Non-kinetic would be the food distribution shown in the counterinsurgency manual or the snatch and grab. And kinetic shaping is basically the euphemism for killing people with tools like that. And I think what we have seen, if you think about it in a doctrinal context and also in the history of aerial warfare is that we have a shift from weapons of mass destructions to weapons of individual destruction. This is basically, they went into Iraq to find weapons of mass destruction. They came out of it with weapons of individual destruction. This is, I think, what has happened, right? So to come to an end, thank you. I think, and this is ongoing research, I'm basically presenting you my toolbox and the ideas I'm developing right now and this, of course, needs to be fleshed out. But what has happened since 2000 maybe a little bit earlier, which at least is what Binny would indicate, what Binny said would indicate is that graphs have emerged as strategic or geo-strategic assets and that there is a race by governmental but also business entities to basically graph out the world who owns the biggest graph, it's important, right? Facebook has 1.2 or 1.4 billion people graphed out, right? Also, there's this problem that they've blurred a distinction between maps and territory because a graph is not just a map, right? It's not just a visual representation, it's actually a dynamic representation of the activity, of the communication, of the relations of the individuals, which you, of course, already shape in the moment you observe it, but which provides a different level of actionable intelligence than a map would do, right? And last but not least, it's really difficult to obfuscate graphs. It's really difficult to get away from them. You can encrypt your communication but you still would be visible as nodes who communicate. And what Binny and other people also Snowden always highlight and that's basically the metadata aspect, that's what graphs do, they ingest metadata and give you an idea who is connected to whom in what form of relations, how close, how far and so forth. So basically it's very difficult to stay out of graphs or to develop tools which will actually create false leads or any kind of other wrong impressions of you in a graph. So to wrap this up, I think we need to think about who owns graphs and what does it mean for us and how can we deal with that? Thank you. We have 20 minutes for questions. So if you have questions, please go to one of the four microphones in the aisles. We're gonna start with a question from the internet while everyone else is getting ready for the microphones. The internet please. We got only one question here. Could you tell us the name of the YouTube movie about human terrain graphics software? Okay, so one is the Palantir demo on counterinsurgency and I actually forgot the name of the human terrain movie. Just Google human terrain and movie, you'll find it. It's on IMDB. I don't know if the whole movie is available on YouTube. Thank you. Okay, then in the front on the, what is that? Left side. First of all, thank you for your talk and I have a question regarding the creation of the grass because if you create those kind of grass, you certainly have some kind of measurement error because it's very hard to create these graphs and most intralid measures are very sensitive to those kind of measurement errors. So my question is, do you know about any kind of research that has a focus on the question? How noisy is this grass and what kind of implications are there because if you make some kind of graphics and maybe you'll kill the wrong guy? Yeah. Well, most of you could. I'm not familiar with the academic literature on that but that's actually questions that, for example, Binny gets asked and doesn't really answer, right? If you look in the NSA-Unter-Zones Alshos, he remains fairly unclear in that context. If you look at, for example, the MIT Network Research Center, you will find papers that address that question because one of the things you will see also in the drone papers by the intercept is that basically they try to either optimize the decision-making so the responsibility can be delegated to a machine or else put it on as many shoulders as possible, right? So there is a review process, of a review process, of a review process involved in those decision-making, decision-making on whom to target and that's how I think they try to deal with that. But the technical side, of course, is probably really complicated. I'm not a technical expert on these questions. Thank you so much. Okay, then question over here. Thank you very much. I just wanted to make sure that I got this shaping thing right because does that mean that people are arrested or even killed just because they happen to be a node in a graph that makes the graph more central when it's removed, not because they did something bad or suspect the terrorist, but just because of the information you can gain from removing that person from the graph? Is that what shaping essentially means? Yes, that's, so I don't know and it's probably without security clearance it's impossible to get the information if greedy fragile, which is the algorithm I showed there, is actually deployed somewhere, right? But the doctrinal papers all basically say you have to identify persons that have a certain influence within the network as depicted by certain centrality measures and either take them to extract information from them or take them in order to see what happens in the graph. But that's exactly what they're doing. That is what shaping does. Okay, thank you. Then we have a question on this side, on the microphone in the front. Good, hi there. Two questions for you. First of all, thanks for presenting this talk. From an American's perspective, it's pretty fascinating to see how this is ingested in a European context. First of all, did you look at the technologies used on mounted on UAVs to pole selectors off the ground and have you looked at the application of biometrics onto these technologies? And then the second question, that's all one question. The second question is about prediction. Palantir and other technologies develop, other similar platforms have predictive elements that are built into their algorithms. Have you looked at how those are used to target individuals within those networks for extraction or whatever? I didn't get the part of the UAVs. Did you study the technologies used to pole selectors off the ground, DRT boxes, so forth? No, I didn't. So I didn't really look into the technical details. So in order to give you an idea what I do basically, I'm interested in the media change of statehood. So how we move from a state that was basically built on paper administrations through digital administrations and how new tools come into the picture in shape or change how a nation state, for example, deals with something like individual identity with borders or the notion of a territory. So I'm looking at that level and I'm not an expert on the technical details here, but look into the SNORN files. There's a couple of catalogs that will give you some information on the tools that they use to acquire the data and capture phone signals and so forth. I don't know anything about biometrics in the context of UAVs. I don't think they're deployed there to a wide extent because biometrical technologies just happen to have too high false positive or false negative identification rates. And I think what they still do is basically have individuals sit there and use humans as tools for identification. Okay, then we go back over here to this microphone, please. Hi, I'm Henrik, I'm a reporter. I've written a number of stories based on classified documents, SNORN stuff. And I can understand your frustration in not having access to a lot of this material. I was wondering, can you talk a little bit more about classified research and which trade-offs you as a researcher see and going into that work, there's stuff going on at Heilbrunn in the UK between the GCHQ and a university, there's West Point in the US, but this is stuff that you as someone publicly facing cannot access. That's the first question. The second one is the relationship between the graph work done in the commercial world and what your impression is of what's going on in the secret world. The strength ratio is one much further than the other one, et cetera. Well, as for the classified research, I mean that's basically something that I have to accept that I work on the basis of leaks and publications by academics in the military context. And one way to deal with that is to basically stay in the historical realm. So look back in time and try to puzzle out what happened in the 80s and 90s. You can't talk much about present-day developments because you just don't know. You can make educated guesses as best. As for the relation between business and government, graphing, I don't know what I think and what is fairly evident if you look at the documents on also academic debates in the first part of 2000, you will find there's a lot of going back and forth. I think that's just, there's not a conspiracy or anything. It's just a discussion, a discourse that emerged where people from academia talk to people from military talk to people in the NSA and other secret services and where these ideas basically were floating around and some people made a business out of it, facing consumers, other made a business out of it, facing the government and others made a business out of it, facing not consumers, but businesses. So I think that's for sure, I'm not, I have no idea how good or bad the graphing capabilities of GHQ or NSA is but it's pretty certain that they use Palantir for example and that Palantir is something like a software package that you can deploy fairly quickly and does what you need. So there's, the commercial sector seems to be pretty strong in this. And over here on the microphone to the left. Hello, thank you for your talk. I've got one more of a remark than a question. You were questioning whether biometric identification actually might be useful. I've got one idea what actually I fear might work and that is movement pattern detection because there definitely is a lot of research also in the public about it and it even works on basis of people that only are like five pixels tall. So yeah, what you will find is talk about pattern of life exploitation and pattern of life is basically mapping out who you relate to but also where you spend time during a day and during night. So what is your bed down place, right? Where do you eat and so forth? And part of this targeting process identification process is basically developing a pattern of life signature for an individual. And that of course is also used for prediction. For example, where will this individual be tomorrow at noon time and so forth. That definitely is part of this process. Other biometrical means like gate detection or facial recognition and so forth. I don't think will be used in drone context. If there's another field is IED, counter IED so improvised explosive devices. They use quite a lot of biometrics there but that's fingerprinting, retina, printing on the ground, scanning on the ground. Yeah, I definitely agree that those things will not be available for drones. I mean, my fear is that things like your walking pattern are just things that you probably have no chance of escaping. True, yeah. Thank you. Quick question in between. Has the internet any questions? The internet is silent. Okay, then we go back over here to the side please. Please speak directly into the microphone. Sure, I have like two quick observations on this. One of them is that there's actually kind of apparently a little open source ecosystem forming. So a few years back, NSA kind of open sourced one of their databases, Acumulo, and that's been picked up by a lot of academia. They've rebranded it, I think D4M, but they're kind of using the NSA code quite happily. And then a few weeks back, GCHQ kind of lay out on top with their own open source database thing. And it's kind of interesting to see that they're actually open sourcing this stuff. The other thing is kind of, I wonder whether you've seen this as well. I think there's a thing about most big data and it seems to apply to these national security things as well, where you talk about it quite a lot, but then in the end, you come down and come down in like day to day business to very simplistic things, right? So you talk about all these graph and behavioral patterns things. And then if you look at the actual documentation for something like X key score, it turns out it's basically a version of graph, right? So it's basically like very, very simplistic matching. And I wonder whether how much of the kind of, have you seen any indications of how much of this is actually kind of, yeah, head in the clouds kind of thinking that then in practice looks completely different. To that question, you know, I don't know because you can't really look inside these projects. But my guess would be it's fairly simple. It's just a scale, right? Which is not simple. It's accessing lots of lots of data points and justing the data and being able to store it indefinitely and search it. Benjamin Bratton, who comes out with a book called The Black Stack, calls this The Big Hall. So basically in the past decade, they could haul in all the data off the world and that provides of course a strategic resource which other entities might not have access to. As for Accumulo, yeah, I found it interesting too. I think basically, I mean, if you look at Hacker News and all those websites, you will find a lot of debates about the software infrastructure of the stacks, of the Big Five companies, Apple, Amazon, Microsoft, Facebook. And I wouldn't wonder if they basically, if there's a lot of knowledge and also tools going back and forth. It's probably, look, if you would have a chart that shows you how the stack looks in the NSA, probably doesn't look much different from Amazon. And maybe Amazon even runs parts of it. And that's what I mean by military, academic, industrial complex, right? There's just knowledge floating around that's know how that is available and that can be employed, you know, academically it can be employed in business context, can be employed in this context. It's just a discourse that emerged once so much data was around and the necessity to graph that out turned out to be valuable. Okay, thank you. And then we have one more question on the left side, please. You told us about shaping of social graphs, the military does, is removal of notes the only operation that was done there, or is there also research on inserting notes or taking over notes? Well, food distribution is not removing notes. It's giving notes something and an opportunity to see where the food ends up, who talks to whom, what do the people tell you when you give them food. And I mean basically targeted ads is shaping, right? I mean if Facebook hands you out something on your timeline, you click on it, it has successfully shaped the graph, right? So, you know, in a military doctrine of context, of course there is an option available that includes removing notes, but the knowledge goes far beyond that, right? The possibilities then just removing the note. Okay, we have a question on this side, on the right side, please. Hi, it's not really a question, just some thoughts regarding the previous question. So, a cumulo is actually a fork of H-Base, which is an open source project. I totally agree that it is to some degree different, but in the end the core is the same technology. And also recently, I noticed that GCHQ has released an open source graph, a two, let's say, distributed graph two. So we'll see. What's the name of it? I have it on my cell phone, sorry. But what I wanted to say is that the industry actually has tools that are really similar, or at least we can think so, to what also the other people are using, okay? Thank you. Thank you. Are there any more questions? Then please go to the microphone now. Internet is silent. Okay, thank you so much for the talk. Thank you.