What satellites reveal about the clash over Kashmir, with Nathan Ruser
24:57 Share

But does it not look right to you as a, you know, as an experienced imagery and geospatial analyst, or would it not look right to me as someone who doesn't spend a lot of time usually looking at these things? It probably depended how closely you looked at it. I think there were some signs that sort of showed that

it didn't look right to anyone sort of like looking super closely at it. But I think the sort of experience of looking at a lot of satellite imagery over a lot of years sort of made a lot more, jumps out a lot more to geospatial analysts. But I think, yeah, there definitely was a case where

a lot of otherwise reputable people sort of got caught up in that. Welcome to Stop the World, the ASPE podcast. I'm Olivia Nelson. And I'm David Rowe. Today we're bringing you another breaking news episode of Stop the World with ASPE's geospatial analyst Nathan Roosa taking us through his assessment of satellite imagery of the recent India-Pakistan clashes over Kashmir. Amid the claims and counterclaims from the two governments and their cheerleaders on social media, Nathan explains what we can determine to be true from satellite pictures and more importantly what appears to be false.

And on the latter, Nathan, who has been among the world's foremost open source researchers using geospatial analysis for several years, is seeing things he has not seen before. We'll let Nathan tell that story as the podcast unfolds. It's a good and timely listen that should be useful to anyone interested in South Asia, disinformation, deepfakes and AI, not to mention nuclear stability. So with that, let's get into it.

Welcome to Stop the World. This is David Rowe and I'm here with Nathan Roosa, ASPE's geospatial analyst. Nathan, thanks for coming on. Hi David, thanks so much for having me. It's great to chat. So the headlines for the past week or so have been dominated by unsettling news of clashes between India and Pakistan. Nathan, you have been doing some of your highly regarded satellite imagery analysis of these clashes.

First of all, just talk us through what you've managed to establish through the satellite imagery about the veracity of the Indian and the Pakistani claims about their respective strikes on each other.

Yeah, so it's been a remarkably difficult conflict to track through any sort of open sources because just the speed in which the claims were coming and the scale of them made it really difficult to sort of track each one and try and verify each one. And I think it's a challenge that I felt and also certainly Indian media felt. I think generally when we talk about the successes and what we can verify from the campaigns...

It's easiest to sort of separate it into two different fronts. Firstly, there's sort of the artillery battles over especially Jammu and Kashmir. But secondly, there's the more strategic Air Force battles between the Indian Air Force and the Pakistani Air Force.

And so when we look at the satellite imagery of Pakistani air bases, we can see that there's quite a number of them, maybe up to 10, that have evidence of quite significant damage through this campaign. And so we've seen satellite imagery of crated runways, of destroyed hangars, destroyed radar sites, and destroyed command and control centers in quite a number of Pakistani air bases spread across the country.

In contrast, India's air defenses have seemed to be a lot more successful in that the Indian government reported that there, I think there were 26 sites that Pakistan tried to strike.

And I haven't seen satellite imagery yet that indicates any of those suffered significant damage. Primarily, yeah, there doesn't seem to be anywhere near the same scale of damage. And it's also, it is possible that because of the Pakistani tactics are more using drones, that the footprint of that damage would be smaller. But yeah, certainly in comparison, there seems to be quite a disparity in the damage that we see in Indian air bases versus Pakistani air bases. Okay. And finally,

For those not following closely the respective statements from the two sides, the Indian and Pakistani governments, is what you've seen actually reflected accurately in what each side is saying about their success or lack thereof? And if not, I mean, are you seeing information operations which are overstating or misleadingly portraying their military successes?

Yeah, so I think Pakistan has certainly overstated some of their successes in terms of the damage that they claim to have inflicted. The Indian government's press conferences seem to be a bit more measured and seem to be a bit more evidenced in that they often will share imagery themselves.

But for example, there's always been that sort of massaging the truth on both ends. Like I think Pakistan claimed to have shot down, I think it was five Indian jets. India sort of originally didn't acknowledge that and then has sort of said losses are part of these operations, but our pilots got back safe. And it appears that the reality is that somewhere probably about two jets were shot down.

So there's always a bit of massaging of the truth. I think one other thing that potentially India could do a better job of is sort of demonstrating why the sites, especially the sites that they chose to bomb in the early stages of the campaign that weren't Pakistani military facilities, were struck. For example, one of the sites that they showed evidence of them striking appears to be a regional health clinic, and the Indian government didn't provide any justification for why that was chosen as a target.

and how they sort of ruled in that it was being used as part of the terror network and not operating as a hospital. And obviously, yeah, I think India could do a better job in sort of explaining the rationale behind a lot of their targeting. Generally, I think that the states haven't been the source of the biggest disinformation, if that makes sense. I think the state press conferences have massaged the truth and have sort of spoken up their victories and spoken down their failures.

But it seems to me as though most of this sort of outright disinformation is coming from sort of these more nationalistic commentators. Right. Tell us more about that then. What are they doing and what are the key narratives that they're trying to promote?

Well, I think the key narrative is just that my side is winning and the other side is in tatters. And I think that comes from the idea that at a lot of points throughout this conflict, we knew that stuff was happening, but there was no information as to what. And so I think a lot of the more nationalistic, national security focused minded blogger community sort of

wanted to fill in those gaps and they amplified rumors they amplified sort of unverified reports and in some cases they sort of made up reports of stuff that was happening and and

We talk a lot about that sort of 24-hour news cycle, and I think in India it's a lot worse than it is here in that the media is just constantly trying to fill information when there was really an information vacuum. So I think that's sort of how a lot of these unverified claims and untrue reports sort of got laundered throughout the community.

Yeah, okay. And tell us more about who these people are. I mean, not obviously you don't know their identities, but who is it? What sort of person is a typical leader? What sort of accounts are they using? And what sort of methods are they using? I mean, are you seeing new techniques being used in this conflict as opposed to some other recent things that you've been watching? Yeah. So I think, again, I think it's worth separating that into sort of the

former military national security commentators that Indian media often relies on. And I think on one end of the spectrum, I think they were often sort of doing a similar thing to the States and that they were talking up victories and sort of trying to present rumors and present unverified facts as more

confirmed, and then that ended up getting reported. Then on the other side there's sort of the social media side, where you've got like a lot of bloggers and a lot of sort of nationalist sort of quote-unquote "OSINT" accounts on Twitter and X, very much talking about how their side is absolutely dominating and how like Jemoo Airfield has been entirely destroyed, and all these various claims that don't seem based in sort of any truth but just seem to sort of either whip up a fervor or sort of get engagement, it's hard to separate.

And so the vast bulk of those claims were really stuff along the lines of just unverified claims saying this has happened. But there were a couple of really interesting incidents where sort of

Not only was old imagery used to sort of, for example, a classic tactic of using imagery from other conflicts or using imagery from other events and misrepresenting that as recent stuff. In fact, I forgot it until just then, but I think an official Pakistani army account put out a screenshot from the Arma 3 video game claiming to show an Indian jet being shot down.

And so that's, yeah. And that's something that a lot of actors have done for a long time. And I think the verification world is sort of used to that and used to verifying those things.

But there was one thing that really struck out to me which was doctored satellite imagery. So there was one account that sort of tried to claim, as I mentioned before, Jammu Air Base had been completely destroyed. And they used screenshots of genuine satellite imagery and photoshopped or doctored sort of impact sites and impact craters onto them. And keep in mind this was sort of in that period where there was no genuine satellite imagery that could be used to verify or disprove it.

I saw the imagery and it immediately was quite obvious to me that it had been fake. Firstly, that's not how damage shows up in the satellite sense that they were using. Secondly, there's a lot of the impacts didn't look right, but I really wanted to prove it was fake by going and finding the exact source imagery. So I made sure I spent a couple of

hours sort of going through all the different satellite imagery to try and find that exact source that was doctored and then you could compare the before and after and basically exactly show what had been doctored. That's something that I've never seen before. I've seen actors try and misrepresent satellite imagery, sort of saying this is satellite imagery of a destroyed hang-up when it's actually from a different conflict.

Or even there was one example where Russia wanted to disprove claims that they had bombed a mosque in Syria and presented satellite imagery to bolster their claims, but the mosque that they had blown up was covered by the title of the slide. So in effect, they'd just hidden the war crime that they'd committed. But this is the first time, at least to my recollection, that we've seen straight up sort of photoshopped doctored satellite imagery to try and

add and embellish damage onto a clean scene, I guess you'd call it. It's really interesting. And I'm

I mean, a few thoughts occur to me out of that. I mean, one is the use of old images. I mean, we've certainly seen that before. I mean, I think of it particularly in Israel-Gaza case. I mean, it was pretty common then. And Israel-Lebanon as well. Broadly across some of the recent Middle East conflicts, that's been pretty common. That Pakistani army example that you used, I mean, it's very hard to believe that that was carelessness. Would you have to presume that that was deliberate misuse and therefore disinformation?

It's a thing that Russia has done a lot in its history, sort of using video game footage. I suspect if you ask Pakistan about it, they would say this was illustrative of what happened rather than sort of saying it was an exact... I think they would deny that it was ever meant to be an exact representation of truth. There was no disclaimer of that. I think very much that is sort of in that sphere of talking up, talking up and trying to make victories out of nothing. So, yeah, I think that was...

It's hard to see that as anything but a deliberate attempt to sort of misrepresent. Just a quick technical question to satisfy my own curiosity. In order to verify that those doctored satellite images were indeed doctored, is there a way of reverse searching those images or did you actually have to go through by hand and compare them? I'm just curious. Yeah, no, it was very much by hand. I could see from just knowing the different satellite centers, I could tell what satellite it came from.

And so it was just a matter of sort of like looking through the recent archive of that satellite. And I think in the end, it was like the second most recent cloud-free imagery that that satellite had taken of that site. So it wasn't hours of guess and checking, but it was very much going, this looks like this satellite, let's go through and try and find the exact scene. And it would have been much harder if they'd sort of gone back and picked imagery from years ago, because then you would have had to

Well, on one hand, you would have had to have gone further. But on the other hand, there would be more changes to sort of demonstrate that it was older imagery. It was a very manual process. And therefore, I mean, that does tell us something about the level of sophistication as well. Although it's an interesting and it's a new technique or tactic that is being used.

As we often find with these sorts of methods, it's not super sophisticated and it's mostly, you know, if you do enough of it, then you start to sort of bend the narrative in your favor. But you know, a little bit like inauthentic Chinese affiliated tweets, for instance, you know, often they're not written very well. And you know, they, to anyone who spends any time looking at them seriously, you can pretty quickly conclude that they're inauthentic.

So the level of sophistication isn't always there. Yeah, absolutely. And in this case, it sort of almost appeared like it was just a nationalist blogger that went open the imagery on Photoshop and then use like generative AI to put the impacts on. So, yeah, it's not a massively sophisticated thing. It's not a massively difficult thing to disprove. It's just a different sort of sphere of disinformation concept.

to what those established sort of verifiers would be used to. So like, I think when you mentioned sort of how much the older or misrepresented images, like I think every,

I think basically every major newsroom now would have a journalist who can reverse search something and do that basic fact checking. And this is just sort of a new iteration of that that I suspect a lot of the existing infrastructure and verification of fact checking might not sort of have these methods to disprove right away. Right.

And what have you managed to establish about the actors or the accounts behind those doctored images? I mean, are they known to be mouthpieces for one side or the other? Are they known as inauthentic actors or have you managed to figure out anything about them? I haven't looked particularly deep into sort of who each of these accounts are. I think, yeah, they mostly just appear to be those sort of nationalistic bloggers that sort of run these accounts.

that basically are trying to whip up like this nationalistic encouragement for their side, if you will. And I think the challenge really sort of comes when there's people that have done genuine work. So like genuine journalists or genuine analysts that can sort of get caught up in a lot of this, especially given the vacuum of sort of actual information. And so we saw a lot of examples of what

generally would sort of be considered reliable sources sort of sharing some of this just because there was so much knowledge that that stuff was happening but so little verified information about exactly what was happening it was sort of this perfect storm for these unverified claims to sort of swirl around interesting so i mean people are probably

accustomed by now to the idea of doctored images, you know, photos. I mean, we were just talking before we started recording about the incredibly lame fake video of Macron purportedly, you know, swiping or Mertz or Macron, I forget who it was, putting a bag of cocaine in his jacket pocket after the idea being it was dropped by Zelensky, who is supposed to be a cocaine user.

completely false by the way, let me hasten to add, but people are at least alive to those sorts of risks. Whereas now with doctored satellite images being a new tool, there's a real lesson obviously for journalists and other people who are actually trying to do the right thing, that they will need to be really, really careful in what they accept as valid and truthful satellite images online. Is that one of the key takeaways here?

Yeah, I think there just has to be, I guess, that within newsrooms and within sort of these verification labs, you

I think there just needs to sort of be an awareness of that this is now sort of an emerging thing. And I think AI particularly, like one of the things that I think has prevented it before is just how hard it is to get sort of a source imagery from above. Like it's, when you think of doctoring, like a photo taken by a phone, it's not hard to find something to put in the place of that. Whereas if you're talking about doctoring satellite imagery, it's a very unique sense, a very unique perspective that's hard to sort of find anywhere.

things before, but with the, with the growth of generative AI, I think it's sort of become a lot more plausible to sort of basically just ask the AI to doctor it for you. And I think, I mean, in this case, there are obvious signs that it wasn't correct from just looking at it. It didn't look right, but it's something that sort of can catch a lot of people unaware. So I think, I think more than anything, it's just like an awareness of how AI can sort of pull into that verification scene and sort of where satellite imagery is, is vulnerable to that.

But does it not look right to you as an experienced imagery and geospatial analyst? Or would it not look right to me as someone who doesn't spend a lot of time usually looking at these things?

It probably depended how closely you looked at it. I think there were some signs that sort of showed that it didn't look right to anyone sort of like looking super closely at it. But I think the sort of experience of looking at a lot of satellite imagery over a lot of years sort of made it a lot more, jumps out a lot more to geospatial analysts. But I think, yeah, there definitely was a case where

a lot of otherwise reputable people sort of got caught up in that. And there's sort of been, I guess, this idea that satellite imagery is almost the ultimate verification in that. And I think a lot of people got caught up in that and just assuming that satellite imagery was true. And yeah, like you said, it wasn't massively sophisticated. It wasn't massively, probably wasn't massively impactful in the end, because even if it wasn't disproved and once it was disproved, the original person deleted it.

once they got caught. But even if it wasn't disproved, it wouldn't take long until new imagery came out disproving it. But yeah, I think it's just something that people need to be aware of. I mean, that's an interesting point that you make. There's a

Well, hitherto there is a sense of authority and trust that people put in satellite imagery as being trustworthy. It comes from a satellite imagery company. As you pointed out at the start, we haven't seen a lot of doctoring of these types of things. I suppose

I mean, going forward, how concerning is it to you as a geospatial analyst if this starts to become more common and you are fighting basically a disinformation battle where you are having to look around and constantly second guess what is being shared out there in public and

use your skills to, I suppose, to discredit or call out to you, obviously, doctored imagery. I mean, what sort of can of worms does that open for someone like you who does this as part of your business? Yeah. I mean, I think for now, in the general sense, what satellite imagery is

most useful for in terms of sort of geolocating and verifying on the ground footage I think that still stays pretty relevant and pretty and pretty authoritative in that you could never sort of ask AI to generate a perfect clone of a scene from above I think that use of verifying footage by geolocating it and even chronolocating it is a relatively safe thing and this seemed to be sort of

Very isolated that sort of battle damage assessment aspect of it. It was sort of purporting to be new imagery when it was doctored old imagery and

compared to sort of when you want to verify whether a massacre took place where it was claimed to, you can sort of match the structures, compare that to even Google Earth imagery going back through time. I think that's a much more solid groundwork in how satellite imagery can still be sort of used as quite authoritative. But yeah, the issue sort of is coming in when there is no new imagery, but new imagery is claimed, I guess, if that makes sense.

Generally, that would be a short-lived thing because it's not that long until new satellite comes out, especially if it sort of gets the narrative. Someone will task an actual satellite to look at what's being claimed and disprove it pretty quick. But it still has that sort of opportunity to shape the narrative until it can be disproved. Yeah. Especially in sort of those areas where, like what we saw in Pakistan, India, where there's a rapidly evolving situation without many verified facts and yet this vacuum of information.

facts but this massive appetite for news and developments and assessments. Okay, just tell us a little bit about the, and we'll wrap up in a moment, but just tell us a little bit about the, I suppose the commercial satellite imagery. I mean where are you, when you do your work, I mean where are you getting these images from and how, I suppose, how much have they started to think about

about some of these risks looking ahead or what's your sense of that? And are there things that are already obvious to you that satellite imagery companies could actually start doing in terms of, I don't know, keeping an eye out for misuse of their imagery or pre-bunking perhaps by, I don't know, like putting out one authoritative image as a starting point just to say this is the real version just in case anyone tries to put out a fake version of it. I mean, are there things that you can see them doing

I guess to some extent we see that in that often satellite companies will release sort of pictures to media of what they have captured. And we even saw that in damage in Pakistan recently where sort of within a few hours of the company acquiring the image, they shared it to media. And so I think that's a good policy. And I think generally they also need to maybe be aware of if the narrative is moving in this direction. And I don't think this particular incident had any real impact on the narrative.

If it does get to their desk, I think they should be willing to put out imagery disproving it pretty quickly. In terms of sort of them themselves having a role in the fact checking, I think it's a lot to ask of a satellite imagery company to do that. And I think it's more in the same way that you sort of wouldn't ask Facebook to verify an image of a war zone posted on Facebook.

So I don't necessarily think there is a role in them in disproving it. Perhaps what actually could be useful, and again, this is top of my head brainstorming, but in the same way that a lot of sort of image generation AIs have some sort of footprint in the metadata that shows this is an AI generated image, I think it would be a thing that it might be worthwhile for these companies to think about in the future if this sort of practice becomes more prevalent in how to sort of

fingerprint their imagery and sort of show that this is original verified imagery. And I think that might in the future, because again, this is an isolated incident that doesn't necessarily have affected the narrative at all. But I think given how sort of the trends in AI and the trends in generative sort of doctoring is going, sort of satellite companies thinking about how they can fingerprint genuine imagery in the metadata or somewhere in the image is probably something that is worth considering.

I mean, that sounds very reasonable. If you're someone like me who follows AI closely, I mean, I assume at this point that everything will be possible with AI sooner than we think. And so I would assume at this point that everything will need to find some other solution that doesn't require any kind of limitations on what can be done with AI. So something like fingerprinting or digital watermarking of some kind seems to me a really

sensible solution that I know, I mean, I'm sure that media companies of all kinds are looking at that. All right, Nathan, look, thanks so much. That's really fascinating. Those who are interested in Nathan's work, be sure to follow him on his social media. He puts out some really, really interesting threads. I think certainly X is one that I follow you on. Nathan, what else are you big on?

Yeah, I think generally I put most of my stuff out on Twitter still. How long that will continue depends to be seen. But for now, yeah, I think check on Twitter as much as possible. I urge our audience to follow Nathan on Twitter if they're not already. Nathan, thanks for your time. Appreciate it. No worries. Thank you so much. It's been a great chat. That's all for today, folks. We'll be back on Friday with our regular weekly episode, which is a big picture chat with one of the world's leading strategic scholars and military historians, Sir Lawrence Friedman. Ciao for now.