Latest / Stage Talks with Bellingcat / "Mapping Militia Networks" with New America
Transcript
- 0:04You're listening to a stage talk titled Mapping
- 0:07Militia Networks with New America. This week
- 0:10we were joined by Candice Rondeau and Ben Dalton
- 0:13from the Future Frontlines Project at New America.
- 0:16They broke down their incredible methodology
- 0:19for tracing some of the most famed Russian militia
- 0:21groups on both Telegram and VKontakte, as well
- 0:25as providing invaluable advice for staying safe
- 0:28online. Whilst doing this intensive work, you
- 0:31can find links to all of the resources mentioned
- 0:34in the talk in the podcast description. This
- 0:37talk was hosted by me, Charlotte Ma, on Thursday
- 0:39the 11th of September 2025, the Bellingcat Discord
- 0:43server. Welcome to this week's stage talk. Today
- 0:58we have the pleasure of hosting Ben Dalton and
- 1:00Candice Rondeau from New America. They both work
- 1:03on New America's feature frontlines projects,
- 1:06Ben as program manager and Candice as senior
- 1:08director. Within the program, they use open source
- 1:11investigative tools, data mining techniques and
- 1:14journalistic methods to explore how network technologies
- 1:16from cyber warfare to artificial intelligence
- 1:19are reshaping global conflict, competition and
- 1:21influence. Today though they are here to talk
- 1:24about their work looking into online networks,
- 1:27specifically tracing online networks of militia
- 1:30and private military groups. You can find a lot
- 1:33about someone online, including who they spend
- 1:35time with both online and offline. This is especially
- 1:39useful for tracking who might be fighting or
- 1:41supporting a group like Wagner for example. Within
- 1:45this discussion we're hoping to unpack what platforms
- 1:47you might find these people in, how you can trace
- 1:50their connections, and what tools might be useful
- 1:52in doing so. As we talk, please make sure to
- 1:55add your questions in the chat box via the message
- 1:58bubble icon in the top right corner of your screen.
- 2:00And please note within your question, if you
- 2:03do not want me to read your username out, as
- 2:05I just said, this is being audio recorded. Okay,
- 2:09Candice Ben, tell us, how do you find these groups?
- 2:15So we were just sort of talking about the connection
- 2:18between Bellingcat and and Future Frontlines
- 2:22and collaboration. I'll just mention that I took
- 2:25a course with Elliot ages ago, Elliot and team,
- 2:30and it really was life -changing for me. So it's
- 2:32really thrilled to be here. I thought I'd just
- 2:35open and tell you a little bit about New America
- 2:37and about Future Frontlines before we kind of
- 2:40dive into the work that we do and talk a little
- 2:43bit about our methodology. For those of you who
- 2:45don't know New America, we are a think tank based
- 2:49in Washington DC. We are non -partisan and non
- 2:53-profit quite genuinely. Unlike a lot of think
- 2:55tanks in DC, we are not a government in waiting.
- 2:59A lot of the people that work here are former
- 3:02journalists or people who are still have one
- 3:05foot in journalism, but also work in policy on
- 3:09specific issues. And Future Frontlines sits within
- 3:13our global programming. We've been around since
- 3:17basically 2018, around the time that I started
- 3:19working on looking at the Wagner Group, which
- 3:22is sort of the jumping off point for the project
- 3:27itself. And I'll give you a little bit of background
- 3:29about that. So, we began this project in 2018.
- 3:36And by that time, The war in Ukraine obviously
- 3:40had been going on for four years. The war in
- 3:43Syria was really heating up with ISIS, very active
- 3:49in Syria, but also in Iraq. And the origin of
- 3:54our work on the Wagner Group began with a question
- 3:57about the changing nature of proxy warfare, the
- 4:01ways in which states in particular, like Russia
- 4:05and other countries, the United States, were
- 4:09engaging proxy actors who were not part of their
- 4:13normal conventional militaries to do their bidding
- 4:18in places in the Middle East and also in Eurasia.
- 4:22And my background is in sort of Russian area
- 4:24of studies and so is Ben's. And I sort of took
- 4:27to the idea that there was a lot that was being
- 4:31talked about in terms of little green men and
- 4:35Wagner. Wagner was just kind of beginning to
- 4:37pick up its sort of brand recognition at that
- 4:40stage in 2018. But nobody really knew much about
- 4:45the organization. You know, it was a lot of rumors,
- 4:47basically. And I think the moment that changed
- 4:51everything for Wagner's profile, there are probably
- 4:55two that we can talk about today. One is the
- 4:59most famous, which is the Battle of Khashom,
- 5:01or sometimes known as the Battle of Conoco, which
- 5:04is when a group of Wagner forces started to try
- 5:10and take control of a gas plant in northeast
- 5:13Syria that used to be owned by Conoco, an American
- 5:16oil company. And it was during this clash on
- 5:19February 7th, 2018 that we saw just tremendous
- 5:25amounts of casualties on the Russian side. They
- 5:28clashed with US special forces who were defending
- 5:31the gas plant in nearby territory. And it was
- 5:35a breach essentially of a informal to formal,
- 5:41what people call a deconfliction line, where
- 5:43US forces and Russian forces operating in Syria
- 5:47and pursuing counter -terrorism operations against
- 5:51ISIS and other players were meant to at least
- 5:55communicate when they were in the same area to
- 5:57avoid these types of friendly fire or kind of
- 6:01semi -friendly fire clashes. But it happened
- 6:04anyway. for reasons we can get into later, mostly
- 6:07sort of backstory politics about what was going
- 6:09on in the Kremlin. And the clash led to about
- 6:14200 Wagner group casualties. And what was interesting
- 6:18about that moment was it really lit up the internet.
- 6:21It lit up the contactia, which is Russia's Facebook
- 6:24and Telegram. There were just wild rumors about
- 6:28how many casualties were actually you know, on
- 6:32both sides. At some point, one of the most prominent
- 6:36nationalists in Russia, his name is Igor Gurkin,
- 6:39who had connections to some of these far -right
- 6:42groups like Wagner, insisted that there had been
- 6:45600 casualties. And so, there was a lot of controversy.
- 6:48And that battle ignited a lot of interest in
- 6:51what Wagner was doing and who it was connected
- 6:54to. And luckily, it coincided with the kickoff
- 6:57of our project on proxy warfare. which we started
- 7:00with two or three primary research questions.
- 7:04One was to try and understand how do PMCs, private
- 7:12military companies, like Wagner actually operate.
- 7:16They were built as a private military company,
- 7:18but they seem to be acting really differently.
- 7:21And they also seem to have connections to other
- 7:26networks of unaffiliated masked men, little green
- 7:31men, including an organization many people now
- 7:35know as Rusich, which has deep roots in St. Petersburg
- 7:38and far -right ultra -nationalist movements.
- 7:42And so we wanted to figure out how are these
- 7:44groups connected. And I was just sort of starting
- 7:47this experiment pretty much on my own. I did
- 7:51have some help from Arizona State University,
- 7:53where I also teach classes in security studies
- 7:57and journalism. And I connected with some young
- 8:02students in computer science who were really
- 8:04interested in also kind of unraveling this mystery.
- 8:07And it was with their help that we began to apply
- 8:10a combination of network analysis and social
- 8:14media analysis to try and understand the footprint
- 8:17of this organization, at least virtually. And
- 8:20through that work, we stumbled on one of the
- 8:23earliest iterations of a Wagner Group channel
- 8:28or follower group, and it was called Cheve K.
- 8:31Wagner Military Review. At the time when we discovered
- 8:35this one single account, there were probably
- 8:39about 6 ,900 followers in the account. And the
- 8:45thing that was interesting about it, as you can
- 8:47see from this picture here on the far right,
- 8:49there's a picture of a statue that looks like,
- 8:53you know, some sort of, you know, a fighter or
- 8:57warrior. And this became, along with some other
- 9:00symbols, a pretty important marker of Wagner's
- 9:04identity because it was actually sitting inside
- 9:07a base in Syria where Wagner forces and Rusich
- 9:11forces were, you know, bunked out in barracks.
- 9:16I didn't know that at the time. I had to do a
- 9:18lot of digging just like everybody else does.
- 9:21and looking at the connections. But there were
- 9:23some things that jumped out of me as a Russian
- 9:26speaker, which was some of the symbology, this
- 9:29sort of throwback to the Soviet era of glory,
- 9:33the use of paramilitary and paratrooper symbols,
- 9:37and also, of course, a lot of Nazi themes throughout.
- 9:41So when we discovered this one singular count
- 9:44in late 2018, it was about 6 ,900. And over four
- 9:50years, those accounts of followers would go to
- 9:53355 ,000. So, that's a 5 ,000 % growth rate.
- 9:58And it was very telling because it really tracked
- 10:01very closely with what we saw was the expansion
- 10:04of Wagner's actual ranks and its operations,
- 10:09not only in Syria, but also a Central African
- 10:11Republic, Sudan, Libya, and eventually, Mali.
- 10:17But it was really a case that kind of floated
- 10:21to the top in some of these chats on this particular
- 10:25channel that really caught my attention. And
- 10:29it was actually when, in 2019, when we began
- 10:34to look into the case of a man who was murdered
- 10:37online, on camera, a man who went by the name
- 10:43of Hamdi Bouta. He's a Syrian. who was born and
- 10:48raised in northeast Syria, had departed to Lebanon
- 10:51to do some day labor work. And in the summer,
- 10:56spring of 2017, he returned to see his family.
- 11:00And upon his return, he was captured crossing
- 11:03the border. And nobody knew at the time, you
- 11:07know, when all this was unfolding, who he was
- 11:10or what his identity was. But the one thing that
- 11:13everybody knew was on this J .V .K. Wagner military
- 11:16review site was a constant stream of conversation
- 11:19about a video that cropped up. It was about two
- 11:23minutes long, depicting several men with masked
- 11:26faces, and you can kind of see them there to
- 11:28the left in the picture, wielding a sledgehammer
- 11:31and beating this man, unarmed man, mercilessly
- 11:35on the ground of some sort of industrial site.
- 11:38And so we used all the usual techniques that
- 11:41you all know very well, geolocation, to kind
- 11:45of try and understand where exactly this event
- 11:48was happening. And of course, OSINT is always
- 11:54a collective enterprise and there was a lot of
- 11:56exchange between Russian journalists and Western
- 12:00journalists who are trying to figure this out.
- 12:01And there was one in particular that bears mentioning
- 12:05is Kirill Mikhailov, who was really one of the
- 12:07first to kind of identify the scene, or at least
- 12:10guess that this was happening somewhere near
- 12:12Palmyra, Syria, where Wagner Trucks had been
- 12:14operating for quite some time. It was when the
- 12:17first two -minute video appeared that there was
- 12:21certainly a sense that this had happened in Syria,
- 12:24but still even then figuring out their identities
- 12:27was tricky. Then, a little while later, another
- 12:33two -minute snippet appeared. And this is a video
- 12:36where the same men were depicted basically beheading
- 12:40and then dismembering the body of the man that
- 12:43they had beaten to death, was a sledgehammer.
- 12:45And then carving the initials – and this was,
- 12:49for me, the biggest clue of their, I suppose,
- 12:53former military unit, the 131st Vedeve. onto
- 12:57the chest of the dismembered body and lighting
- 13:00it on fire. All of these things were kind of,
- 13:04you know, the regular practice of Wagner. In
- 13:08some ways, a lot of torture videos, a lot of
- 13:10trophy videos was a way for them to bond and
- 13:13also to kind of show their prowess to each other,
- 13:17but oftentimes the material was leaked, basically.
- 13:20We sort of were able to pull everything together
- 13:24through a couple of different techniques. With
- 13:26the help of our student team at ASU, we tweaked
- 13:30a tool that we found online on GitHub that was
- 13:35very useful for sounding out a geodata and metadata
- 13:42posted on Vakantaktia. And Ben is showing here
- 13:46actually. our tool, the Future Frontlines Ghost
- 13:50Tracker, where if you enter a particular location
- 13:55and a time and a date and you drop a pin, you'll
- 14:00usually find within a 50 -meter radius, if anybody's
- 14:04posted pictures in those areas, you can usually
- 14:06find those pictures. And it was through a series
- 14:10of soundings around the Palmira area. that we
- 14:14were able to locate a number of people who were
- 14:19posting pictures that had a very similar sort
- 14:22of background to what we were seeing in the background
- 14:25of that video of Hamdi Bhutto's murder. And it
- 14:30occurred to me that actually, while we could
- 14:35kind of see pieces of the puzzle, we couldn't
- 14:37see the whole picture. But when we sort of pulled
- 14:41back a little bit, and started to look at the
- 14:44networked relationships between the people who
- 14:47were posting pictures and the networked relationships,
- 14:50people who were friends of certain users in the
- 14:54CHVK Wagner account who were extremely active.
- 14:58It's then that we realized actually the network
- 15:00effects of this organization are much more powerful
- 15:03and easier to trace. And it's a way to sort of
- 15:07additionally verify connections between people.
- 15:10Of course, there's more work to do and then we'll
- 15:11talk a little bit more about what that looks
- 15:13like. But this picture that you're looking at
- 15:17here, this image, represents a slice of a network
- 15:21that we carved out of both looking at the JPEG
- 15:25metadata from posts using our ghost tracker on
- 15:30the contactia and then also looking at the friend
- 15:35relationships between different individuals who
- 15:39had signed up to J .V .K. Wagner Review and had
- 15:43certain profile attributes, many of whom were
- 15:48based in St. Petersburg, many of whom had served
- 15:51in the airborne VDV forces, many of whom said
- 15:56that they had served in Donbass in 2014 and 2015,
- 16:02several of whom bore Nazi logos or Rodin -Avare
- 16:06symbols, which was the Russian sort of pagan
- 16:09religion. And at the center of this network,
- 16:14it's kind of hard to see from this level, but
- 16:16there were three or four individuals that really
- 16:19stood out to us as A, closely connected. So we
- 16:23were really also tracing the tightness of the
- 16:27network and the closeness of the relationships.
- 16:30And we discovered Oddly, at the center of that
- 16:35were some of the most active members of Russia's
- 16:37ultra -nationalist movements, including a guy
- 16:41named Alexei Miltakov, who is now quite famous.
- 16:43He was the commander of Rusich, this ultra -right,
- 16:47ultra -nationalist neo -Nazi group. And then
- 16:50several members of the Russian imperial movement
- 16:52and the Russian imperial legion, which is an
- 16:56incredibly important part of the Wagner group's
- 16:59kind of background. as many of the individuals
- 17:02who trained with Wagner in the early days had
- 17:05paramilitary training in St. Petersburg through
- 17:09the Russian imperial movement, which was founded
- 17:12by members of the Vedeve airborne. So there's
- 17:15a lot of common threads. And if you sort of using
- 17:20network analysis techniques and big data analysis,
- 17:25we were able to kind of see how they all connected
- 17:27to each other. What was interesting about this
- 17:29slide, I'll just mention this before handing
- 17:33it over to Ben to talk a little bit about more
- 17:35of what we found, is that four individuals on
- 17:39the right, so Milchakov and the three others
- 17:41from Russian Imperial movement, had actually
- 17:43been sanctioned by this time, by the time we're
- 17:45doing this work. But the one on the left, Yashakov,
- 17:49had not been sanctioned. He was actually quite
- 17:51active and very important to some other operational
- 17:57parts of how Wagner worked that we later discovered.
- 18:01So what we understood is basically, if you want
- 18:04to really understand how these kind of shadow
- 18:07operators work, nine times out of 10, you're
- 18:10going to find them talking to each other online
- 18:13somewhere. And that's no mystery to most of you
- 18:15who do this work. But for us, what was revelatory
- 18:18was that the contact you in particular had some
- 18:21interesting design features that made it extremely
- 18:24exploitable. Most importantly, because the Russian
- 18:27state seems to be very keen that online platforms
- 18:32like Vakantaktia maintain metadata for surveillance,
- 18:37it makes it easier for other people, including
- 18:40public interest investigators, to also mine that
- 18:43metadata and put the big puzzle pieces together
- 18:46using network analysis. Over to you, Ben. Thank
- 18:49you, Candice. So, yeah, the approach that Candice
- 18:53was describing was working so well for us that
- 18:55we decided to expand it and continue it over
- 18:58a period of years. So there were two main waves
- 19:01of collection. So just to recap, you know, we
- 19:04have identified these VK groups of interest using
- 19:10the tools that Candice was describing, using
- 19:12tools like the Ghost Tracker to find accounts
- 19:14of interest. And there was a period of collection
- 19:16that lasted from 2019 to 2020. And then again,
- 19:20we did another round of collection in 2022, about
- 19:23six months after the full -scale invasion. And
- 19:26we focused really on three overlapping groups
- 19:28on BK. So one was that CHVK Wagner Military Review
- 19:32Group that Candice was talking about. The other
- 19:34was the main kind of official group for Rusage.
- 19:37And the third was the main official group for
- 19:39the Russian Imperial movement, all of which had
- 19:41overlapping memberships, as you can see if you're
- 19:45listening live on this chart here. And just to
- 19:49be clear, by collection, what I mean is that
- 19:51we expanded to bulk data collection of members
- 19:54of VK groups using the VK API. So this enabled
- 19:59us to create datasets with full profile data
- 20:01for every member of these VK groups. And that
- 20:05included their intergroup friendship connections.
- 20:09I'm told that this is an audience that appreciates
- 20:11some degree of technical detail. So I will note
- 20:14that between these two periods, we had to make
- 20:16some pretty significant changes to the API script
- 20:18to accommodate changes that VK itself was making
- 20:22to their API that made it a little bit more difficult
- 20:24to do this. And we also had to contend with the
- 20:27groups themselves practicing sort of greater
- 20:29and greater degrees of OPSEC and sort of information
- 20:32security as the years went on. But because of
- 20:35these two main periods of collection, we were
- 20:37able to compare how they changed over time. Candice
- 20:40already mentioned this huge group in the Wagner
- 20:42group, but pretty much they all experienced significant
- 20:46group over time with a significant membership
- 20:50core that remained substantially the same across
- 20:54the years. So as I said, the VK API gave us public
- 20:58data for members in these groups, and that enabled
- 21:01us to do a pretty granular analysis of graphics
- 21:05and the geographic distribution of the members,
- 21:08or at least as self -reported on their account
- 21:11profiles. And what I'm showing now, I'll describe
- 21:16it for those listening after the fact, it's a
- 21:18map of just one of those variables, which is
- 21:20specifically self -reported locations. As you
- 21:24might expect, the majority of them are in former
- 21:26Soviet countries, but there is just this massive
- 21:28geographic distribution. I think we found 177
- 21:32countries in all. around 5 ,000 of those people
- 21:36based in Europe. And in the interest of time,
- 21:39I won't show you other variables that was military
- 21:42unit identification. There's a field on BK where
- 21:44you can say where, you know, in what military
- 21:47unit you served. We hired a team of researchers
- 21:50to basically identify which of those were real
- 21:53existing military units, which were historical
- 21:55and which were fictional. People would claim
- 21:57to be part of Warhammer 40K and like fictional
- 22:01units. And then we mapped the ones that are actually
- 22:04existing. And long story short, we were able
- 22:06to identify some active duty members or service
- 22:10members of units in NATO countries using that
- 22:13method. So a couple of slides ago, Candice was
- 22:18showing you one network analysis that showed
- 22:20interlinkages as some of the central figures
- 22:23in these groups. What I'm now showing is sort
- 22:26of looks like a tangled ball of yarn. So this
- 22:29is one of our working network visualizations
- 22:32of membership links among those three main VK
- 22:35groups that we analyzed in 2022, which yielded
- 22:39some pretty interesting results. So we were able
- 22:42to identify highly central nodes in this network,
- 22:45even if they were not a member of any of those
- 22:47three groups. So in other words, there were VK
- 22:50members who had extensive friendship ties to
- 22:53members of these three groups, but were not themselves
- 22:55members. which led us to dig really deeper into
- 23:00some targeted profiles and yielded some interesting
- 23:03insights. And I will again shout out the work
- 23:06of the Information Competition Lab at Arizona
- 23:08State University, who we've worked with for many
- 23:11years on subjects like this and has helped us
- 23:14a lot with this kind of analysis. So when you
- 23:17put all of this together, you have data from
- 23:20figures that we identified through social media
- 23:22collection and network analysis. We had sort
- 23:26of a more typical OSINT collection of other public
- 23:29data across the web. You had public bloggers,
- 23:33such as these two gentlemen who are now deceased,
- 23:36who would often share useful information, whether
- 23:39or not they realized it. And this led us to identify
- 23:42figures of interest who we believed played a
- 23:44role within the recruiting and command structure
- 23:47and propaganda structure. of this organization,
- 23:51specifically the Wagner Group, but also a larger
- 23:53network of Russian paramilitaries as well. So
- 23:56I joined the team in 2021, and a couple of years
- 24:01after that, we gained access to a cache of internal
- 24:04data that belonged to companies associated with
- 24:06the Afghani pre -koshen. And much of what we
- 24:09found at that point corroborated our earlier
- 24:11work in terms of who we had identified that was
- 24:14of interest. So, you know, after collecting everything
- 24:17that we had, We organized it temporally, geographically.
- 24:22We ended up building robust dossiers, we called
- 24:25them, of key figures within this network. We
- 24:28selected people who were prominent or held command
- 24:32positions, but also potentially for their alleged
- 24:35involvement in atrocity crimes. and published
- 24:38a set of these, about 24 of these, about a year
- 24:41ago, fall of 2024. And we're continuing to work
- 24:44on more. The slide that I'm showing you now is
- 24:47one of my, I don't know if I would say favorites,
- 24:50but a gentleman named Vladimir Katayev, former
- 24:54Spetsnaz GRU, who, you know, I think he's interesting
- 24:59because he really shows the ability to move up
- 25:02the ranks. So he was like I think a platoon commander
- 25:07of some kind in 2017. But by the time that he
- 25:11was participating in the assault on Bakhmut in
- 25:152023, he was one of the assault attachment commanders.
- 25:19Is there anything that you'd like to add to what
- 25:21I just said, Candice? Yeah. Well, one thing I
- 25:24think is interesting, I'm so glad that you picked
- 25:26the Kataev picture because the through line here
- 25:29is actually Kataev was also one of the perpetrators.
- 25:32in the murder of Hamdi Bhutta. And it was his
- 25:35unit. And we all discovered this later. And that
- 25:38was the most fantastic part about, you know,
- 25:40sort of doing this work over several years with
- 25:43different researchers kind of coming and going
- 25:45and helping us, but continually applying the
- 25:47same methodology, which is, you know, big data
- 25:50collection, small sort of anecdotal OSINT verification
- 25:55combined with network analysis, those three things.
- 25:58Plus, a pretty robust knowledge of sort of how
- 26:02the Russian military is organized, what its traditions
- 26:04are, and sort of digging into that helped us
- 26:08to kind of actually see this through line from
- 26:10one crime. And, you know, nobody knew that Katayev
- 26:14was actually involved. But it was later when
- 26:16we also got access to some of these leaked documents
- 26:19that were shared with us from our partners, C4ADS,
- 26:23another organization based here in DC, that we
- 26:25discovered actually there was a whole set of
- 26:28reports in which Kataev and several other members
- 26:32of his unit were identified as the perpetrators.
- 26:35So, it was both gratifying and also kind of scary,
- 26:39you know, the level of accuracy that we were
- 26:43able to kind of produce by combining all these
- 26:44methods. was really quite powerful. And if you
- 26:49go online and you take a look at our site, which
- 26:53is still active and we're still trying to find
- 26:56ways to upload more information there, you'll
- 26:59see actually not only were we able to do a network
- 27:03analysis and identify individuals, we were also
- 27:07able to really put together with all of these
- 27:09details, the command structure, which when you're
- 27:12looking at war crimes, is extremely critical
- 27:15for understanding who was giving orders when
- 27:18atrocities occurred in a given area. In many
- 27:21cases, through triangulating all this data, we
- 27:25were able to also establish who was where when
- 27:29and which units. And so the dossiers are a reflection
- 27:32of a small batch, but in actual fact, we have
- 27:35a rather large database of about 13 ,000 personnel
- 27:39whose rank and their movements and their background.
- 27:43has been all documented. The one thing I just
- 27:47need to say is Bellingcat has been part of this
- 27:50journey along the way. We've always had them
- 27:53as a sounding board, along with our colleagues
- 27:55at ASU and C4ADS. As always, OSINT is a collective
- 28:00action power that we all have to tap into and
- 28:05rely on to get the work done. Over to you, Charlie.
- 28:07Thank you so much for giving us a lowdown on
- 28:10your methodology and how you have looked into
- 28:13this group. There's a thousand questions in the
- 28:19chat, which I'm really excited to speak to you
- 28:23about. Tristan actually said when we were unpacking
- 28:26a lot of what you've been doing, collecting information
- 28:29like this and platforms is always a cat and mouse
- 28:32game. Both the platforms and the targets change
- 28:34their behavior and the collection has to respond
- 28:36in kind. One of the biggest questions that's
- 28:39been coming out is how do you make sure that
- 28:43the people that you are IDing, how do you make
- 28:47sure that you haven't got any false herrings?
- 28:51Chris asked, what percentage of profiles mapped
- 28:54do you expect are red herrings? How do you confirm
- 28:58those IDs? Yeah, well, a lot of work has gone
- 29:01into the dossiers in particular, I will say,
- 29:04and they're I guess there's a distinction and
- 29:07Ben can talk a little bit about kind of the mechanics
- 29:10of doing transforming raw data from leak files
- 29:16that shows a pattern of command structure versus
- 29:21looking at where people were when. So, when we
- 29:26take the dossiers, what we're really doing is
- 29:28kind of scissoring down like through the network,
- 29:31almost like taking a piece of the spider web
- 29:33and then putting it under microscope. and saying,
- 29:36OK, what do we know about this individual that's
- 29:40in the public sphere? So we do all the normal
- 29:42things. We look for corporate registration information,
- 29:47tax IDs, matching birth dates. We look for their
- 29:53own social media profiles, which oftentimes many
- 29:56of them were still live. Or interestingly, Not
- 30:01surprisingly, the community of OSINT, who's very
- 30:03interested in the Wagner Group, had spent a lot
- 30:05of time archiving also some of these live profiles
- 30:08that were very popular and well known. And in
- 30:13addition to that, medals were also handed out
- 30:15to individuals in many cases from the Kremlin.
- 30:18And so we could also check sort of the official
- 30:21register to see, you know, which medals were
- 30:24handed out when. And actually We had a researcher
- 30:29from Stanford University, a fantastic statistician
- 30:32and data analyst who helped us kind of build
- 30:35a verification for our data set. And Ben will
- 30:38talk a little bit more about the technical aspects.
- 30:41Yeah. So I should note that things that are just
- 30:47purely social media data, we would not... published
- 30:50personally identifying information without extensive
- 30:52corroboration from other sources. So when I showed
- 30:55that map earlier that showed the geographic distribution
- 30:58of members from these groups of interest, that
- 31:02was not identifying who they were individually.
- 31:04It was only identifying the geographic locations.
- 31:07Whereas when we identify a specific person, like
- 31:10the Kitaya figure, that has been extensively
- 31:12corroborated with, obviously, their online activities,
- 31:15but also internal documentation. And by the time
- 31:19that actually we were, we published this, he
- 31:20was already essentially a quasi public figure
- 31:22because there were like news reports about him.
- 31:25Um, in terms of, yeah. So Ken has mentioned we
- 31:28have this, um, this dataset of Wagner group personnel,
- 31:31um, going from all the way back to 2014 up to,
- 31:36uh, I think it's 2022. Um, and for that, you
- 31:41know, this is basically, uh, you know, diskewing
- 31:44their own internal personnel lists into a highly
- 31:48organized and structured data set. And so the
- 31:52data is coming directly from companies that are
- 31:54associated with or were associated with the Gepgini
- 31:57precaution. And that's the corroboration that
- 32:02we're looking for and that we've gotten. But
- 32:04even in that case, we've not made that data set
- 32:05public. So it's not freely available on the internet,
- 32:09because again, it contains an enormous amount
- 32:11of personally identifying information. The people
- 32:13who we explicitly identify publicly are like
- 32:16sort of the crème de la crème in terms of cooperation
- 32:19and involvement in these activities. Absolutely.
- 32:23And cooperation is super important when it comes
- 32:26to identification. As you mentioned, there's
- 32:29been a couple of people asking about what you
- 32:31do if you find that when you're tracking someone
- 32:35across profiles, that they've got different usernames
- 32:38on different platforms. It's very common. Often
- 32:41people spell their name differently as well.
- 32:44What kind of techniques have you found across
- 32:47platforms from this particular group when they're
- 32:51trying to maybe hide who they are or obscure
- 32:56the ability to trace them? Being a member of
- 32:59a paramilitary organization or any organization
- 33:03that systematically uses violence for enhancement
- 33:09of their own influence. Even in secret organizations,
- 33:14there's a lot of flexing within their own networks.
- 33:18So, by that I mean there's kind of like a lot
- 33:21of symbology, there's a lot of bragging, there's
- 33:23a lot of, you know, seemingly secret chatting.
- 33:28I'll just give you one example. So, Rusich is
- 33:31a fantastic example. We have a colleague who
- 33:33was extremely obsessed, probably still is extremely
- 33:35obsessed with Rusich. and following the Instagram
- 33:39account of Rusich, which was live up until about
- 33:432019 -2021, actually, and sort of had gone up
- 33:49and down, but was like the most active way to
- 33:51see what was going on with Rusich. And one of
- 33:53the habits that they had was putting information
- 33:57in kind of a coded chat to each other, sometimes
- 34:01using, you know, poetry from the Poetic Edda,
- 34:05which is this sort of Viking era poetic epic.
- 34:11But then also using slang that actually most
- 34:14of the slang comes from military culture and
- 34:18particularly very specific military culture.
- 34:22Just like the special forces in the United States
- 34:24have a certain way of talking, so do the airborne
- 34:26forces of Russia. And the vast majority of the
- 34:29most active and most kind of high profile members
- 34:33of Wagner and Rusich came from the Airborne Forces
- 34:36or Spetsnaz or both. And that's stuff that you
- 34:40just have to kind of study. But one thing that
- 34:44happens oftentimes is because they're so enmeshed
- 34:47in that culture, even when they want to separate
- 34:50themselves by like having different aliases,
- 34:53almost always there's some sort of weird signature
- 34:56overlap. You know, because a lot of them are
- 34:59Nazis, just as an example, you might see the
- 35:01number 88 come up quite a few times, right? You
- 35:05know, there are other things that are just such,
- 35:07you know, if you spend too much time in these
- 35:10far -right circles, you kind of start to recognize
- 35:12them and you can see the links between them.
- 35:15And then again, I just want to, you know, commend
- 35:17all the people out there who are doing such a
- 35:19good job of archiving a lot of this material.
- 35:22And the other places that we don't talk about
- 35:24very much but are kind of important to recognize
- 35:27are the kind of weird dead bases of the internet,
- 35:31like .su. which is the Soviet Union, the old,
- 35:35you know, domain area for the Soviet Union. That's
- 35:38where a lot of far right, ultra nationalist conversation
- 35:41goes on. And that's where you can do a lot of
- 35:43verification of how people change their identities
- 35:47over time based on the kind of symbology that's
- 35:49been used there. Probably Ben has some other
- 35:52tips too. Well, yeah, I mean, and this might
- 35:54seem a little bit obvious, but like, so on both
- 35:57VK and Telegram, which are the two still the
- 35:58two main platforms that I think that we monitor
- 36:00most closely, These groups, they have a dual
- 36:05purpose, where they're public -facing in the
- 36:09sense that they serve a propaganda function,
- 36:11they serve a recruitment function, sometimes
- 36:13they're being used for crowdfunding. But at the
- 36:16same time, they're important for internal purposes,
- 36:20to form a sort of internal culture and internal
- 36:22coordination. And so they're public, but they're
- 36:26quasi -public, right? Um, and because of that,
- 36:29we also understand they all exist on like the
- 36:31knife edge being banned at any moment. Um, and
- 36:34so they will often have one or more backup groups
- 36:37that they will fall back to if their account
- 36:39is banned. And, uh, if you just like get really
- 36:43obsessive and spend an enormous amount of time
- 36:45in these communities, you, you effectively, you
- 36:47know, you'll have as much information about their
- 36:49online structure as the, as the members themselves
- 36:52do such that, so just to pick one example, um,
- 36:55We've been talking a lot about research. They
- 36:57had a telegram channel that I'm going to get
- 37:00the dates wrong because it's been a couple of
- 37:01years now. But back in the fall of, I want to
- 37:04say 2023, it was banned because they posted explicit
- 37:08instructions for like how to torture prisoners
- 37:10of war and then extort their family members for
- 37:12Bitcoin by like, you know, holding the location
- 37:18of where the body was buried. And this was too
- 37:20much even for telegram. And within 24 hours,
- 37:23the channel was banned. Um, but they had, you
- 37:26know, usage too, that they had already created
- 37:28that stood up, um, immediately. And so they,
- 37:30you know, operations continued right back at
- 37:32it. And, um, that's maybe not the best example
- 37:34because research is like pretty, you know, prominent
- 37:36on telegram. It's not hard to find them, but
- 37:38like, it's a, it's a, it's a consistent pattern.
- 37:40Um, and if you just are really obsessed and spend
- 37:43a huge amount of time watching them, you will
- 37:45be able to know about all of their little backup
- 37:47and, you know, monitoring can just like continue
- 37:50as seamlessly as their operations do. We often,
- 37:52whenever we're talking about network analysis
- 37:55and particularly identifying individual people,
- 37:58often we speak about the ethics of maybe infiltrating
- 38:02kind of closed groups or what's the difference
- 38:06between monitoring afar and then also kind of
- 38:10friending or befriending people on platforms
- 38:13so that you can monitor their profiles using
- 38:15a sock puppet. How do you kind of balance those
- 38:18ethical arguments within your research? Yeah,
- 38:22well, I mean, I think everybody wrestles with
- 38:24those things. And I think as the field has evolved,
- 38:27and also platforms have kind of evolved, as we've
- 38:30been talking about, you have to evolve your own
- 38:33ethics, of course. But generally speaking, you
- 38:35know, of course, I always had a lot of young
- 38:37students who were, you know, helping with the
- 38:39research. People were much more computer savvy
- 38:42than I am, who wanted to kind of engage with
- 38:45these guys. And I had to explain, you can look,
- 38:49but don't touch. And that was like kind of a
- 38:52rule of thumb for all of us, largely to protect
- 38:55us in our own security, because this is an organization
- 39:00or sort of a network of organizations that is
- 39:03very bent on vengeance and targeting people.
- 39:07And of course, they're connected, broadly speaking,
- 39:10they were connected to Yevgeny Prigozhin's various
- 39:14media enterprises and propaganda enterprises,
- 39:17most notably the Internet Research Agency. which,
- 39:19you know, has a legendary hacking capacity and
- 39:23trolling capacity. And still to this day, actively
- 39:26does a lot of trolling of individuals who study
- 39:29and work on the Wagner Group or look at, you
- 39:32know, some Russian propaganda and ultranationalism.
- 39:34So, for us, it was kind of a hardbound rule that,
- 39:37you know, we were not to engage. Of course, we
- 39:39use sock puppets, you know, to constantly do
- 39:42our monitoring. And we were very regular about
- 39:46it. In terms of... And we never did any sort
- 39:49of hacking, you know, there was no, that's also
- 39:52a no -go zone for us. We want to stay within
- 39:55the bounds of the law and sort of ethical standards
- 39:58as best we can so that the work can actually
- 40:00do the good it's meant to do, which is to expose
- 40:03perpetrators of war crimes and atrocity violence.
- 40:07I guess the follow -up question there, as Chris
- 40:09has just asked in the chat, is what are your
- 40:11top tips for OPSEC? What are the... steps that
- 40:15you take to protect yourselves? Well, they're
- 40:16the basics. I mean, obviously, you know, if you're
- 40:19going to be monitoring on social media, always
- 40:21better to do it with sock puppet. Good to rotate
- 40:24your sock puppets as well. Keep them active and
- 40:27alive and looking like they're engaged so they
- 40:30don't look like sort of zombies that are sort
- 40:32of trawling around on the platform. Try and make
- 40:34them to some degree fit with the culture that
- 40:37you're looking at. So you can sort of camouflage
- 40:40yourself. You shouldn't be sort of, you know,
- 40:43wearing maybe like a rainbow flag if you're entering
- 40:46an ultra -nationalist right -wing space. As an
- 40:48example, right, that shouldn't be your sock puppet
- 40:50identity because that will immediately attract
- 40:52attention. So trying to blend in is really important.
- 40:57Using a VPN, critical, must do it all the time.
- 41:00We use burner phones, as many people do, for
- 41:04a lot of our interactions. So as an example,
- 41:06I don't have Telegram or VKontakte on my phone.
- 41:10I usually have another phone for that so that
- 41:12I don't have any crossover between my personal
- 41:14or professional life and the research and investigation
- 41:18I'm doing. And we, you know, we don't really
- 41:22talk about what we're doing until after we're
- 41:23done doing it. And so we oftentimes work with,
- 41:26you know, sometimes as many as, you know, 12,
- 41:3015 different sort of student researchers or faculty
- 41:33researchers or people that we're sort of in collaboration
- 41:36with. And I think, you know, we've done a relatively
- 41:39pretty good job of sort of minding our P's and
- 41:42Q's when it comes to just being quiet about the
- 41:45work that we're doing. No point in bragging about
- 41:47it because there's nothing to show then. So,
- 41:50and in fact, this isn't really sort of an enterprise,
- 41:54I think, that really is good for bragging. I
- 41:56think there's just, you know, the work requires
- 41:58a certain amount of humbleness because you can
- 42:00also make a mistake. And I think you also have
- 42:02to recognize that on some level. I'm sure Ben
- 42:04has some other tips too. Yeah, I mean we also
- 42:07in addition to burner friends we've used for
- 42:10burner air gap laptops Occasionally over the
- 42:13years who've gotten I guess you could call like
- 42:15external hard drives of dubious origin That you
- 42:18don't want to just like plug into your regular
- 42:20device for pretty obvious reasons And so that
- 42:21we have a sort of sacrificial lamb laptop that
- 42:25you could use for that This another one more
- 42:29thing that I think speaks to a question like
- 42:31one or two questions ago Which is that if we're
- 42:34publishing like a major report? um, on a, on
- 42:36a public figure, um, we will, you know, do traditional
- 42:41journalistic diligence and, and ask for, um,
- 42:44comment. And that usually does not, you know,
- 42:48usually they're not interested in speaking to
- 42:49us, but there is a stage right before one of
- 42:52the last stages, uh, before going live where
- 42:55we will do that. Right. So that's sort of part
- 42:58of the, the basic ethics of of doing an investigation.
- 43:01Yeah. Just circling back to kind of techniques
- 43:03that the groups also use to try and obscure their
- 43:06identities. Somebody mentioned earlier, you know,
- 43:10how Vucic seemingly intentionally time -lags
- 43:13location -based content posted to its Telegram
- 43:15channel. Does that happen on VK as well? I think
- 43:18in those cases, chronolocation becomes a skill
- 43:21that is absolutely paramount. It takes time and
- 43:25effort. Um, but that's where you don't just rely
- 43:28on the time tag of when something was posted,
- 43:31you're constantly checking, um, if you can do
- 43:33locate and then chrono locate particular imagery
- 43:36or videos, for example. Um, I don't know if you
- 43:38have anything else to add to that particular
- 43:40point before we move on in terms of ways that
- 43:43people obscure information on mine. Um, yeah,
- 43:45I'll just say a few words. Uh, but research absolutely
- 43:48does this all the time. Uh, they're kind of notorious.
- 43:51They have seemingly just like an inexhaustible
- 43:52archive. Um, that they will post from all the
- 43:56way back to like, you know, here's Milchakov
- 43:59in Syria in 2017, or here's where we were in
- 44:04Ukraine circa, you know, late 2022. And they'll
- 44:08often actually identify it with like the location
- 44:09and time, which, you know, you can't really take
- 44:12their word for. Or, you know, alternately, they'll
- 44:15post without any kind of identifying information.
- 44:19I think that in general, they're pretty careful
- 44:22and pretty good about not revealing anything.
- 44:24Um, that would give away their current operations
- 44:26or at least their operator, you know, in anything
- 44:27that could affect their ongoing operations. Um,
- 44:32just to, yeah, just to, you know, I'll agree
- 44:34with the point that you made that you have to
- 44:35do a lot of due diligence to actually confirm
- 44:38that the thing that they're showing you, um,
- 44:41is from the time and place that they're claiming
- 44:43that it's from. Yeah. Don't just take their word
- 44:46for it. Circling back to the tool that you mentioned,
- 44:50we've had a lot of questions on that. A lot of
- 44:52people are interested in it, as you can imagine.
- 44:55Is it open source? Is it available anywhere?
- 44:59There's been a lot of people searching for it
- 45:01on GitHub, for example. Can you tell us a little
- 45:04bit about the development of the tool? If it's
- 45:07open source, where is it? And if it's not, how
- 45:11come and is there a plan to do so? I knew this
- 45:14question would come. I do believe the actual
- 45:18original code. is somewhere in GitHub. And perhaps
- 45:23after the fact, we might find a way to share
- 45:26that. Our code is updated and cleaned and refined
- 45:31for location accuracy. We haven't put it online
- 45:35or made it open source yet. Again, for some of
- 45:39the same reasons that we're a little bit sort
- 45:41of nervous about sharing some of the data that
- 45:43we have. We're trying to sort of wind our way,
- 45:47and I think we probably will find a way to make
- 45:50a lot of our tools, techniques, and data more
- 45:54accessible, likely on a tiered basis. Again,
- 45:57because this is really sensitive information
- 45:59that we are hoping will inform war crimes investigations
- 46:04that are ongoing. Not only in Ukraine, I just
- 46:06want to mention there are places in the world
- 46:08where Wagner and Rusich have operated where there
- 46:12aren't a lot of human rights defenders out there.
- 46:15So Mali is a great example. Syria is another
- 46:17good example. And yet there is, I think, an appetite
- 46:21still to see some justice brought. And I guess
- 46:24that's why we're a little bit cautious about
- 46:26being too open with our tools and methods and
- 46:30data. However, in certain settings, we are happy
- 46:34to share with certain... types of folks who are
- 46:36kind of interested in that particular type of
- 46:38work. So yeah, the version, the original code
- 46:41does exist somewhere on GitHub in the pre -modified
- 46:47version, but in order for it to work functionally
- 46:52today, you'd have to make pretty extensive changes
- 46:54to it, both the ones that we made, but also literally
- 46:57just to like update it to work with VK's existing
- 47:00modern API, which changes all the time. The first
- 47:03version was just a week or two. We had some very,
- 47:07very clever mathematicians and computer scientists
- 47:11on our team who managed to tweak it. But Ben
- 47:15has progressively tweaked it again and again
- 47:17as the API has changed, as platform rules have
- 47:21changed. So it's one of those tools that has
- 47:23to be constantly updated. Gio Crow has just asked,
- 47:26what is the name of the GitHub project? I'm guessing
- 47:28you don't know from the original source code.
- 47:31I'm sure we have it somewhere. If we can find
- 47:34a way to share it, we will. Fabulous. Is there
- 47:36any tools that you would recommend outside of
- 47:39the ghost track tool that kind of do similar
- 47:43functions or are useful at least for searching
- 47:46telegram connections or VK connections? So I
- 47:51have this dream. This doesn't exist yet as far
- 47:53as I can tell, but it's something that I would
- 47:55love to see built. which is something that would
- 47:59do something similar for Telegram. Now, you can't
- 48:02build something that just like searches for geotagged
- 48:04content on Telegram the same way, because almost
- 48:07nobody posts geotagged content on Telegram. But
- 48:11one idea I had was, especially in parts of the
- 48:14world that are relatively less dense and more
- 48:18remote, like for example, a relatively undeveloped
- 48:22part of Mali or the Central African Republic,
- 48:24a tool that like the ghost tracker that we showed,
- 48:27draws a bounding box around a geographic area
- 48:30and then searches for recent posts that mention
- 48:33community names, drawing from like open street
- 48:37map data within that area in a variety of languages.
- 48:42So like Russian, English, Arabic, et cetera,
- 48:45which could be a way of sort of like fishing
- 48:47around for interesting posts and identifying
- 48:50channels on telegram that you might not have
- 48:52been aware of. previously that are posting content
- 48:56on that. And if anybody out there has the technical
- 49:00chops to build something like that, I would love
- 49:01to see your work. Yeah, other tools I would just
- 49:06mention, and these are things that actually,
- 49:08I mean, I'm pretty sure our methods are replicable
- 49:11because we've had to reiterate them over and
- 49:13over. But I certainly think if you're scraping
- 49:17data, particularly social media accounts, but
- 49:20I would say exclusively, I would say there are
- 49:22other kinds of data, too, where you can create
- 49:26a searchability and connectivity capacity there.
- 49:29So doing network analysis, for instance, on social
- 49:32ties or common characteristics and attributes.
- 49:35But also, again, the Information Competition
- 49:39Lab has been extremely helpful. Some of the work
- 49:41that we did, for instance, on the January 6th
- 49:44Parler data that was released very shortly after
- 49:47the riots on Capitol Hill here in DC, was really
- 49:53facilitated by the creation of essentially a
- 49:57SQL database that allows us to dump open data
- 50:01into it and then search it by text and number
- 50:07and moniker. Those are things that actually can
- 50:11be replicated, I think, with relative ease, especially
- 50:13now with the support of AI. And again, if we
- 50:17get time, and some sort of support out there
- 50:21in the world to continue doing our work. We could
- 50:24probably come back and kind of share some, just
- 50:26some tips on specifically how to do that. And
- 50:28maybe invite our colleagues from the ASU lab
- 50:31to join us. That sounds amazing. Saiva has put
- 50:35in a really good PSA though in the chat. Fair
- 50:37warning to any budding code developers. Any trolling
- 50:41tool has a high chance of leaving a significant
- 50:43footprint on the platform you're accessing. which
- 50:46means your activity could be tracked back to
- 50:48you and used maliciously. So please, please be
- 50:50careful if you are building tools to do such
- 50:55a job. Yeah, 100 % on that. I just want to say
- 50:59the tools I'm mentioning, especially, you know,
- 51:02kind of these specialized search tools for particular
- 51:06data, all of that's happening in a virtual machine.
- 51:09So really important to work in a virtual machine
- 51:11or in a virtual environment that is shielded
- 51:13and not connected to your target. Absolutely.
- 51:18I also noted in the chat, if you're interested
- 51:20in this specific discussion, you might want to
- 51:23check out our previous discussion with the All
- 51:25Eyes on Wagner team. They talked about monitoring
- 51:29Wagner mercenaries in Africa. So you can look
- 51:33at that conversation, if you would like, on our
- 51:36RSS feed. or on any podcast platform by searching
- 51:40StageShorts with Bell and Cat. We've only got
- 51:41five minutes left, which is a shame because there's
- 51:45so many more questions in the chat. Guys asked
- 51:48earlier, are there any best practices you can
- 51:51identify in regards to extracting an accurate,
- 51:54incredible pattern of life from web data, especially
- 51:57in the context of the platform and community?
- 52:00I mean, I think it's really specific to the community.
- 52:03I have to say that This is such a specialized
- 52:08community, and yet the signature of how they
- 52:12kind of tend to act and to flex and to communicate
- 52:16does have a lot of transferability. So for instance,
- 52:21we've looked at others of American far -right
- 52:23groups that have kind of similar ways of communicating
- 52:27with each other. And I guess as a result of a
- 52:30lot of our work, we've also inherited a lot of
- 52:33data that people want to share with us. because
- 52:36I think they kind of have a sense that maybe
- 52:38with combining our methods, we can kind of, you
- 52:41know, get a sense of the pattern of life. The
- 52:44pattern of life is super hard. I don't even think,
- 52:48you know, some of these AI programs like Lavender
- 52:50or whatever that the Israelis are using. I mean,
- 52:53they're just... they're always going to be dirty
- 52:56because the data is always going to be dirty.
- 52:58So you shouldn't be expecting like 90 % accuracy.
- 53:02You're just always going to have to triangulate
- 53:04to understand pattern of life, which is mostly
- 53:08really familiarizing yourself with the history
- 53:10and culture of the social group that you're looking
- 53:14at. You can't just go on faith of kind of leaked
- 53:17corporate data and then some social media and
- 53:19then that's it. Because you really need to understand
- 53:22what is it that glues this this network of people
- 53:25together. And people work in networks. The only
- 53:28biggest lesson is every social group in the world
- 53:33operates in a network format. And it tends to
- 53:38be, you know, the more extreme, I suppose, in
- 53:45kind of commitments to violence, the more likely
- 53:48the signaling amongst that network is going to
- 53:50become super obvious and quite routine. and very
- 53:54easily recognizable. I don't know if Ben has
- 53:56any other observations. Yeah, I don't know. A
- 54:01word of caution, I guess, which is that at this
- 54:03point, you know, I assume that most of the accounts
- 54:07that we track are probably using a sort of a
- 54:11hybrid pattern of sort of generative content
- 54:15and then human generated content. And it will
- 54:18switch back and forth between them. I think most
- 54:20platforms at this point have gotten I don't know,
- 54:24relatively good at detecting just like purely
- 54:26automated content and you just like the pure
- 54:29bot. Um, and, uh, but at the same time, if you're
- 54:32just like a single human plugging away during
- 54:34your sort of information operations, then you're,
- 54:37you're not, you know, competing at the level
- 54:40that you need to in today's information ecosystem.
- 54:43And so I, yeah, I, my, my hypothesis dealing
- 54:47with these, these groups and entities is that
- 54:49it's sort of like a hybridized. pattern, which
- 54:52makes, yeah, pattern of life really difficult
- 54:54because you can't really tell when their online
- 54:57activity is sort of the human or whatever sort
- 55:01of generative content is also potentially posting
- 55:04on that account. What tipped you off for high
- 55:07level command profiles identifying the command
- 55:10structure? What was the key signifier that you'd
- 55:14found someone who had a high level command? So,
- 55:17this really started with that Chebyka -Wagner
- 55:20review group that we mentioned on Vkontaktia
- 55:22that we kind of stumbled on in 2018, just as
- 55:26all that stuff was happening with the Battle
- 55:27of Khashom and their world was kind of lighting
- 55:30up. One thing that happened on the anniversary
- 55:33of that, so in February 2019, so one year later,
- 55:38was that somebody, and I still theorize to this
- 55:42day that it was probably a Ukrainian off. But
- 55:45somebody threw up a message saying, you know,
- 55:48I remember, you know, the fallen of the Battle
- 55:52of Khashim, I'm paraphrasing here, you know,
- 55:55and then the next chat line was, who served where?
- 56:00And weirdly, at that moment, just suddenly like
- 56:06dozens and then dozens and then hundreds of guys
- 56:11would just... answer with their Vechay number,
- 56:15which is their military unit number. And I realized,
- 56:21holy shit, these guys are just openly identifying
- 56:25themselves. And I'll note that that's when I
- 56:27also realized that many user accounts on Bacontactia,
- 56:31you know, especially for guys, it's pretty typical
- 56:33to have the unit that you served in as commscript
- 56:37posted. It's kind of like where you went to high
- 56:39school, where you went to college, university.
- 56:42It's part of your profile. You don't have to
- 56:44fill it out, but many guys do because it's a
- 56:47way to find each other and also to show that
- 56:49you've done your service, you've been a patriot
- 56:51and so forth. And so we collected all of those,
- 56:54put them into a spreadsheet, and then we did
- 56:56a lot of verification. This is the early days
- 56:59even before Ben was around. You know, our first
- 57:02attempt at this was kind of the data was a little
- 57:04bit dirty and we didn't understand at first how
- 57:07many were imitating units and how many were real.
- 57:11It was only later as we started to refine our
- 57:13methods and realize, oh, we can actually segment
- 57:15this profile data on military unit affiliation.
- 57:21But that was the most telling piece was just
- 57:24where people had served and then having an understanding
- 57:27of who was central to the network. And the vast
- 57:30majority of having a particular profile and a
- 57:33particular affiliation with units that were known
- 57:36to be very active, for instance, in Georgia in
- 57:392008. and some other sort of ghost operations
- 57:42that well preceded the incursion in Ukraine.
- 57:45So a lot of it, again, is this combination of
- 57:48kind of understanding the culture, like the literal
- 57:50culture, not the online culture, and then doing
- 57:53a lot of historical research to put it all together.
- 57:56Amazing. Thank you so much for giving us that
- 57:58detail. I know Chris, who asked the question
- 58:00initially, really appreciates it. Thank you so
- 58:04much for coming into chat today. I have a bunch
- 58:08of people in the chat right now saying how much
- 58:10you've humbled them and also how much they've
- 58:13enjoyed this talk. Please do check out both Candice
- 58:16and Ben's work at New America. I put the link
- 58:19in the chat and I'll also put it in the description
- 58:21of this podcast, along with all of the links
- 58:23that we've discussed today. And please tune in
- 58:26in two weeks time for the next stage talk. But
- 58:29until then, take care and thank you again, Ben
- 58:32and Candice, for your time today. Thanks so much.
- 58:37Thanks, all. Thank you, everybody. Thank you
- 58:41for listening to the Stage Talk. If you'd like
- 58:44to catch a Stage Talk live where you can ask
- 58:47the guest questions, join the Bellingcat Discord
- 58:50server by visiting www .discord .gg The music
- 58:57you've heard is titled Dawn by Newer Self and
- 59:01is courtesy of Artlist.