***Please note that this is an automatically generated transcript and will contain errors.***

---

So continuing a direction of the enforcement, with respect to actually knowing where the peers are running, where in this type, in this line of work, you're focusing not only on having km wide precision about ideally being able to pinpoint the particular confidential virtual machine you use, that particular, data center, which comes with a different set of requirements that I would be talking about.

And one is like very fresh, direction that we've also been working with, with this of the work of space computer interaction. How can we use a something like basically a structure to provide the required information for the approval operation as well? So the first line of work is focusing on the other use cases. That I would say is a bit more mature.

And the motivation for it actually is let's just like the different isn't necessarily focusing on the aspect of your this optimization with rather on how to strengthen the, the ease of that model. But a second one I think can also go in the direction of, strengthened data control, but also provide additional guarantees. So the general aspect as we try to focus on is to provide some form of like a either computation binding, to the respective execution environment, which is the work of the data center execution assurance.

And the second one is the leverage, some of the aspect of communication binding, of knowing that a certain PR is running in a particular location. So similar discussed before with using something like a challenge response. See how this can be actually used in in different context? Well, of course there are certain legal challenges that we have to resolve.
Of course, as I think we might use operators, there can be some network instabilities, and to see what can be done in this part of. So starting with the first work, it's focusing on the other, solution where we actually are trying to hardened aspect of these confidential virtual machines, that are running in certain aspects of the infrastructure where the reason for that is for our little bit certain execution environments.

We are facing some challenges with respect to physical outputs, how this can be resolved. So we choose to rely on some aspects of the operator not trying to physically tamper with the device itself. So then if if you look into the details of the, the station doesn't provide any information about, location until it's running for the provider either.

So we are looking to the way essentially how can we find these two information together, which eventually basically gives the guarantees of that a particular potential virtual machine is running in, for example, Google Cloud, infrastructure, due to the fact that we are able to buy, confidential. If you have some additional information from from the provider, which the motivation here is mainly on, how can we ensure that the these we are running integrating with is basically more secure where of course, if you have high value transaction, and you are deploying your system before cloud inside of the T, you want to make sure that, is the case.

But of course, it is also leads to the aspect of getting the confidence level to location price actually executed. And and for that, we are we already published initial run is plan motivate this particular problem. And we are currently working on a solution that will be hopefully made on archive later on. This month that is focused on both things, actually, of confidential virtual machines that are running inside the cloud environment, but also on bare metal assumptions that are also running in the environment.

So the motivation here would be that we are looking for ways and how can we by bind, cloud provider details to the first execution like so. And the way that we envision to do it is Google actually owned virtual trust. But for you, that is part of, several deployments of providers that is able to later on correlate the information that are coming from the environment of the station, together with the, Vpm, other station fault and can leave it on providing the guarantees of the particular, location with your virtual system.

So that part we necessarily care, about the other external activities, but we know that the particular platform is running in particular datacenter, actually, and we are rather looking to the other way on how to inject some privacy into the into the system to make sure that this type of sensitive information about its own, a region of your reporting of workloads are not leaking, but just providing information that is running in a trusted environment.

By the fact of relying on the providing. So, and we have a solution, hopefully for both bare metal deployments in cold provider settings, but also for conventional virtual machines, which allows us to have additional, trust and, confidential information that we are interacting with them on that voice of the workloads in them. Of course, if you consider it on a large scale, I guess it's the basic amenities everyone would have to rely on these, which, depending on the setting you're in, this event might not be completely fitting.

And the second direction is very early detection. So bear with us. It's something very fresh, that we are working on also in the crowd, so it's more like an open input from community. The would be very happy to, to possibly work on together where for those also that do not know us from space computer, we are trying to be considerate of the performance of those in space, and we are looking into the ways in how can you provide some sort of services through just on Earth.

And the idea that we were based amongst, the, the team was to look into some solutions, for example, like gold, or similar that are actually, in solve the problem of robust cloud storage on Earth, but try to extend it by leveraging satellite infrastructure. So, as was pointing out previously, you will need some, assumption on the hardware.

This may be available on the device, on the terminal on Earth, because otherwise GPUs and other solutions are very, very easily. So can we do, something without the or to be always inseparable is on Earth and basically prevent it? Or can we just design a general purpose protocol when we are interacting with a bit of use on Earth and maybe argued it also used physical location, and then we are connected and part of the here they have to of course it and reach the satellite.

And this is to provide us some additional guarantees as well. So there are, I think, few open questions, that in general before us, we know from the part of typical satellite communication in the context of GPUs, as we know. But this is something that will require, for example, terminals running on the, the infrastructure itself, which might even get you suitable for applications like data centers, because if you look at it, many of the data centers nowadays are already relying on running a dedicated, network interface cards, for example, for solutions like systems like AutoML or PPE, but actually giving very precise clocks from space based infrastructure to be able to synchronize, their own clocks using the protocol for which they actually have a dedicated hardware already in their infrastructure. So then you just need to extend it by additional aspects of, additional aspects of the GPS location with a certain triangulation to a particular area. And then the cloud providers themselves should provide it for the individual viewers and the system.

So I don't know, for example, whether it's using this area optimally in their infrastructure already, because they of course are interested, for example, for lobbying when certain informations are coming in, but it's also being used quite actively enjoyed by with respect to data centers as well. So there is like is there a way on how to extend on this type of assumptions, to, to give a better understanding or the other way around?

There's already a lot of current infrastructure for space based communication, either relays are the type of infrastructure that is around the world, and you can try to map it out and see basically what are the latencies between the individual, base stations and gateways on the from later on, how to triangulate very something like location, whether or with the respective of the on earth that will be orchestrated through the information that we are gathering.
So this was a lightning bolt. It's nothing that it's for the first install. And I just wanted to provide some possible direction where we have different set of assumptions on the computation of solutions we are relying heavily on, of course, the usage as the confidentiality. But also the VPN, providing us the route of trust coming from the provider.
But of course, in case we want to have a robust antenna resistant solutions for the satellite infrastructure, we also have to consider some form of, suitable capabilities by the provider that can maybe be tried to spoof the information that are coming out of the, the respective devices. Or we have maybe put some trust, in the, in the respective infrastructure provider.
But maybe if you find actually an interesting direction, that could be something.

An interesting question we are currently working on is actually is there some data that there is basically a proving location on Earth captured by the prove the satellites in space that will inherently also be able to bridge the gaps between the location, or because you have to be able to communicate with the respective point in time that actually can disclose the information about your particular location.

So I think, there are interesting direction that could be pursued for both of the computation, but also for the communication one. Yeah, that's pretty much. And thank you very much.
Questions?

Yes. So for the data centers basically you will use the. And then I would just like on the roof and like click on like special device. Then it would be even possible. So I think you know the first I quote in their communication or extension bit by respective data floating to the provider or it's small idea maybe to the provider itself.

It just doesn't provide this information as a service to the individual years. And they can say the information. You mean information like the time like for from GPS. Exactly. So okay, so this is for example the current setting for the Milan protocol. You have a dedicated network interface card for which you have a dubious receiver. And then internally inside of the data center they have infrastructure that is BGP capable, meaning that you are able to synchronize circles in an event position and are distributing this, quick information to the peers in your system.

Well, the first, it assumes that the fields are honest, right, in this context. But since it's usually provided to you on the host operating system level, and maybe it's like even like a single provider, you can have enough assumptions on the front or the first. If you go into the managed, the scenarios where the parties can just delay the responses whenever they want to.
But it's.
Run then your slides. It's a can be spoofed. Does this I can you can you guys reconcile all that? Seems like there's, I just very guessing this feels like GPS, so. And I'm curious, so, maybe you can spoof GP's as. So Vegas works as it's, bank starts to be a sort of x rays supplied sensor.

And then you say calculate your dealing based on fortnight size, like for instance, to determine your own time and your location. Right. You just artificially manipulate the sensor. So you captured earlier or later in time and then release it to the receiver after 17. And then that's how useful, because you just be like, explicit connection.

Yeah. So basically here you don't be relying on timestamp, sort of on the GP itself, like it's the protocol because as mentioned, like it doesn't provide too much security retrospectively, but you would have to be using maybe our infrastructure or our like infrastructure with respect to providing confidential. Then we of course would assume that there would be some form of trusted element, possibly on the device itself.

So you're replacing GPS with like a T on both sides of the executor. Yeah. So you can have a physical layer in between. Why do you need, t on the ground wouldn't then be sufficient just to have trees and space doing the triangulation, but then the peer itself can just do whatever they want with, that information, everything.

They could try it in a modified a time synthetically. Right. Or post, you can measure the response time back to this, to the sample so that that doesn't what you call just one by then you will get I think you get some guarantees low. It'll be still either is it is is the upper bound right. You would have to.

Yeah. It would have to consider I think the something that you're relying on, you will get some information. But then I think silly. So I think in this case would be the maybe I was just like using something like a VPN to run England as I'm in this particular location challenge. So if you want to get at a confidence of your particular vertical distance from at this particular location.

So you could also like pull in proofs from like different places and like that simple like, yeah, say like, okay, this is my GPS. Yes. But you've collected time stamps at like six months close and forwarded them. So you need to have also some confidence I think in the themselves there's a slight yeah. Yeah. Other question. If this fails it's kind of a troll, but like a consistent what's my struggle?
What about putting one in a volcano? To the question okay. Let's go, let's go. Yeah, yeah, I might need to revoke the the squeeze on the data, but it really limited the machine for the user. Well, there's also one in carbon fiber. So right next to the Titanic. Yeah. Very robust.
Yeah. No, it's not able to do it. Yeah, yeah. But okay. Actually serious question about, implementation. So whatever you're using to launch the satellite. Yeah. So this is just today, you know, and I'm sure you're launching your on, can they control your device or not? Like, is this like, and they like a hypervisor to your machine.

You get them to launch in the space. You mean like they just introduce other like, so. So yeah, I mean, there are different models. So current book is becoming more popular. And, it's like basically software as a service. So you can just like read the comments or you can also like still own the aspect of infrastructure and something like that so that you control the host operating system, VM, whatever you want as well.

So there are there are different and different solutions, different settings. Okay. You can also have like a multi-tenancy. Basically it's like the net was like a code base deployment. So you could just like run your virtual machine easily on something like that space. So for example, companies like Start Code or others are actually trying to provide this low environment.
Okay. Yeah. Yeah. So just wondering to what extent like software side channels. Yeah. It's a bit of definitely for that I think you need the bar for if you look at a layered approach of having the liquid key on the side, I would solve it will be leveraging desktop the physical isolation system in the sense of several bucks person.

So, I really this question of security. So you might know what what's the law about code space. Who our knowledge currently I think there is not much it available regarding this. Yeah just Slovenian. So the green signal captured in orbit from Earth, it's, you have two stories and, it is, you know, it's you're emitting is back to Earth.
You have to storage for just so in years. I think you can of deleted. So to be very choosy about what you present back to Earth. This is a big problem for companies like planet that maps the entire Earth. But most of their pictures are just clouds. If they don't seize for them, if they say vector, is that to keep this excellence of the data, I don't know that there's any regulation regarding code.