A New Frontier in Confidential Computing SpaceTEEs and L2 Security - Filip Rezabek

Transcript

[Music] Test test perfect just oh spoiling all the fun let me go to the intro slide yes so hello everyone uh my name is Phillip I'm currently as nicely introduced pursuing my PhD at the technical university of Munich in the area of distributed computing and in parallel I am also a co-founder of space computer and today I'll be presenting aspects on how can we actually send execution environments or in short tees to space and identify what additional aspects it can provide to the users on earth and how we envision the future frontier with respect to confidential computing in space.

So I have several points to cover today. In general, of course, because we are a rather a young project, uh we also have to take the responsibility of educating our audience with respect to why do we foresee that the space is the next frontier and why today is the good timing to actually try to send and build a distributed system in space. Then we'll dive into the aspects on what is actually the temperroof infrastructure we aim to achieve. provide an overview of our 2T architecture that many of you might be already familiar with from the Ethereum ecosystem and try to apply it to the context of layer 1 solutions running in space. Of course, highlight briefly the smart contract capabilities, dive into some of the challenges that are unique for our L1 but also L2 and provide aspects on actually how space

computer can help with our current solutions we have on earth to make them more secure and robust. So firstly, why do we actually currently see the push of and good timing to actually build a new solution in space? So the data speak for themselves. In the last couple of years, we have seen launched more objects to space than in the past decades. And we also see more importantly that the cost of launching any amount of uh satellites or especially the payload sizes is becoming cheaper over time. Meaning that in the maybe 10 years ago before SpaceX was a thing, you would have to pay hundreds of thousands of euros for each kilogram of the of payload you are launching.

Nowadays the cost is becoming part of it. Meaning we are talking about thousands of euros and of course with the competition that we currently foresee either from Space X, Blue Origin, you name it. The payload cost should be only decreasing. What does it mean to us? It means that now is the good timing and we can see it on a regular basis when you look into the news that more uh satellites are being launched to provide confidential computing or internal type of computing in space. Maybe some of you registered two weeks ago uh China has launched a new complete constellation of satellites for AI running in space. We have projects like loft and many others that are building fewful data centers in space where the main promise of the technology in space is to directly collect all of the energy from the sun

without the need of relying on a power supply from from earth itself. So definitely this is something where we envision the long-term future and of course uh due to the aspect that you are in space uh you also can actually improve on the latency not as of now as it is but of course moving forward due to the fact that you are in a much better environment where the data can propagate with much less delay itself and more importantly as many of you maybe are already users of such technologies uh or have devices equipped with it we start to see that from traditional solutions such as Iridium there start to be broadly available solution ions such as Starling that of course will enable maybe not as of now but in the future much more performant communication among not only to users to

satellites but also uh intra satellite communication while they are running in space. So what is space computer? Uh our target we have as of now is to build a new layer 1 solution in space. Meaning that actually we really want to run L1 among the satellites running consensus among satellites and in addition to that we are aiming to each satellite itself providing confidential computing capabilities meaning that each of the satellite we called space TE or space execution environment. I will dive into it in upcoming slides in more detail.

what what is the unique aspect of the TE we of course uh as many other elevants currently are providing we envision the future where actually you need to have programmability capabilities in space and therefore of course we want to be part of it with smart contract platform capabilities u and of course as of now we envision several use cases you can think of it like a co-processors for tees but of course we see the foreseeable future that there will be more and more computation and orchestration of the computation happening in space for which we want to provide a robust marketplace and payment layer uh among other satellite providers as well. Of course uh we currently have just uh published a blue paper last year in November and we are currently working on white paper with more specific use

cases and application that I'll be happy to discuss with you during the runtime of the conference as well. Looking at the architecture briefly the idea is I have to be pointing somewhere over there I believe maybe it's going to work or not doesn't seem to be the case. Uh so you have of course the earth components and around that we will have the individual satellites that will be forming together the distributed system with which we call of course space computer and then you will have additional satellite service providers that will be orchestrating their payment systems and others across our platform itself. So what does space space computer provide to users on earth and why should you possibly consider using uh our solution? So in general uh we envision that we will have individual

satellites providing capabilities to its users but of course depending on the on the applications you can also have the full-fledged aspects of uh actually distributed system capabilities. So you can think of it uh that you can just communicate with a single satellite. It will be providing trusted execution environment capabilities like for example you are used to with interacting with Intel TDX or others just with a slightly different and hopefully improved threat model. You can for example by interacting with a single satellite uh collect uh CT RNG or as we call it cosmic through a number generator or you can have a custom wallet solution also stored on the individual satellites. So this is just the aspect of a single node solution.

But of course overall what we envision is that you will just have the seamless experience as you are used to from interacting with your uh blockchain of of your favorite uh uh L1 through RPC nodes have dedicated communication links that will be able to communicate with your peers and uh have in general additional aspects. We don't only only envision ourselves that we'll be communicating with users which is of course important for us but also want to provide uh robust solutions to other L1s for example with respect to secure bulletin board aspects or checkpointing where you can think of it that since the infrastructure is running in space it's very hard to be tempered with and therefore you can think of it like a additional settlement or uh robust layer where actually Peters can also store

that information from earth itself. looking into the aspect on why are we motivating to build uh space computer as it is as of now following uh example of Ethereum in this case to have a two-tier architecture uh as you might have guessed we are kind of constrained when it comes to building something in space either on both communication but also on the computation of course so for that in order to scale we have to look into the ways on how can we achieve it and for that fortunately there is a lot of huge body of research already done that we can look into and actually apply it to ourselves as well. with of course the additional constraints, communications and other parts that will enable scalability and usability to users while

actually also providing uh different type of guarantees. For example, robust communication among the peers since when you think of it the space satellites they are available pretty much from anywhere because you have them always orbiting over you and you don't have to rely on the underlying physical infrastructure on Earth itself. As long as you for example have ID or Starling terminal, we envision that you will be able to communicate directly with our infrastructure. Of course, since we see that uh the computation and communication capabilities of space uh satellites will be improving, we also have to consider aspects of dynamic upgrades upgrades over the consentus mechanism itself which is something that we are now currently actively working on on how to make it into reality with the work on the white paper itself.

So overall uh what where we see our aspects is that we will have a very robust uh L1 that will be running in space and we will have L2 solutions for scalability itself. So uh this is something where we also try to provide some novel approach that of course as we know for example from optimistic or zerops we want to a little bit push it forward in combination with state channels uh to also provide additional usability uh guarantees itself we are highlighting this part in the bluepr I don't have time to dive into it now at the ACE in in more detail but um just let's continue farther with where do we actually provide the resilience on two levels I was talking about communication and computation so let's dive into the aspects of communication bit more in detail. As of now, we are relying on the

Iridium network. Uh some of you might already know it with respect to NB watches. You name it, when you are in the middle of nowhere, you are still able to communicate. So, this is the current state-of-the-art basically solution that is providing to the satellites as well capabilities to provide intra satellite communication and for that itself is orbiting already around us with roughly uh 70 satellites enabling us to provide a good coverage over the earth. We are currently partnering because we see ourselves as a protocol company, software company uh with various types of hardware providers that are actually providing us the access to the compute itself. But of course we envision in the future that we can expand on that and we basically are trying to combine both of these

capabilities together to build a robust protocol that is having unique hardware capabilities and of course uh it's providing most importantly resistance to type of denial of service attacks because uh that the infrastructure itself is uh providing this type of capabilities. So this is something that is uh very I think unique sort of feature. You can think of it like also uh as a fallback mechanism as well to current systems on earth where maybe you are just like used to uh rely on your uh internet service provider. But of course in case you want you can also have a dedicated link that is slower maybe more expensive uh but serves as a good backup solution for your type of communication via the radio network itself.

The second flow is uh the focus of I would say the today's talk mainly which is the computational resilience that we focus on the aspect of space te. Uh so this concept is already introduced in 2017 by the previous co-founders of uh at that point cryptosat orbit called nowadays that introduce the general idea of physical isolation due to the fact that you're running uh your satellite in space. It's very hard to go and temper with it because it's rather uh difficult to get a hold of it because of the location right and also actually the speed as well. So even though the currently there start to be more and more aspect of advertis activities in space still it's rather challenging but of course we don't want to stop there we

want to expand on it further which I'll be talking on the upcoming slides but for those who are not familiar with the concept of TE you can think of it like a dedicated isolated region on a CPU that is providing additional instructions uh to the users when it comes to building uh building the solution and it aims to provide a secure location for your both code and data itself and in general It aims to offer confidential integrity and also depending on the solution you're using it aims to support attestation capabilities meaning u that this is I think very crucial feature that we are aiming to to have as well of course and the the aspect of remote stationation is that I can convince you or anyone who wants to interact with the infrastructure that actually your code your data are properly being executed

inside of this secure environment and you can do this through various approaches and there are also we will talk about it on the upcoming slides with a different route of trust but the general idea is that you are confident that your transaction your confidential transaction or you you can uh think of it as a general compute has been executed inside of the TE. the additional capabilities that it also is providing to provide transparency uh to its users because you actually with the other stationation itself. You also know for example what type of binary or other informations are running inside of the TE as well. Meaning you can for example add a state the guest operating system possible also the host operating system depending on the model the binaries and other parts that of course if you have

later on open source uh the individual components others can take recompile it and verify the information that actually they are interacting with doesn't have for example any back door and with that you are building the root of trust and the aspect of interaction with such solutions u maybe over the last past years you have heard especially I think last year it was becoming really a hot techn technology uh just providing a high level overview of what are the various types of TE's out there. Uh so in general there are two main categories process-based TEES and VM based TE's. Uh for the process based solutions uh the general idea that is that you as a host operating system are running uh your doing your regular operation you are

running your binary and just like a single part of the process are basically executed inside of the TE. Uh some of you might already have heard of Intel SGX or many of you actually might have it on your phone is the ARM trone. uh and then we have the other solutions which I would say are mainly enabling uh currently the usability coming with security which are the virtual machine based uh solutions such as Intel TDX, AMDSVSMP and especially I think interesting for us is the ARMCA technology uh that is basically building on top of the trust zone aspects but allows for additional privacy and additional isolation and here we stand is the aspect of space te which is providing physical isolation but of We want to provide the software isolation

as well. So you can think of it that we are layering different type of T technologies together u where we try to have similar or possibly smaller uh trusted compute base as the VMbased solution but want to of course build on already established solutions that are out there uh that we can actually later on uh launch inside of the on the satellite itself. So briefly speaking we can think of it we for example take arm trust zone expand on the threat model of of arm trust zone due to the physical isolation and other aspects of the verifiability of the individual software stack of on the individual satellite to achieve much better security guarantees that you would have received uh by running the same technology on earth as

well. I didn't mention it here on the slide. Uh but of course for those that are in the TE space I think a very hot technology we have seen also recently for example in the Ethereum ecosystem is the focus on risk 5 and this is actually something that we are very actively looking into as well because for example um there is a keystone project that is enabling a full-fledged open source capabilities that actually can be uh used and deployed as well. The very interesting part of risk and ARM where they are the main focus points for us is due to the fact that we are computation constraints that is one part. So we are usually running ARMbased architecture in the future possibly a risk 5 based architectures but also we are power constrainted meaning that we cannot just

take any of the shelf server that will be running Intel TDX and run it in space. We are also constrained on the power budget itself because of the cooling uh of the equipment we have as well. And of course the more power we we would need the more solar arrays we would need which of course make the whole payload much heavier. So we have to balance uh among all of these capabilities to achieve uh what we are aiming to achieve uh with the robust infrastructure itself. And now kind of coming into the part of that we outlined what are some of the features we can offer with respect to computational and communication resilience. Let's look into how can we possibly collaborate with various L2 solutions that uh the first one we cannot really help with but

when you look at the ecosystem and some of the discussions there are known bugs uh in the implementations of either the smart contracts or actually the proving systems as well that currently are very hard to be exploited due to the fact that it's rather central centralized solution right we uh are aware of that as a community but it's something that it's still worth uh pursuing to to improve and I think this is actually what is actively happening and for that events like this are very important where the soundness of the implementation of uh has to be improved and validated. We also start to see more shift uh towards uh instead of single prover models to multip models meaning that you are for example combining TE together with zero knowledge rollups as

they are that are actually providing additional incentives to the operators and their users. We currently also know that there is certain level of trust that is added into the sequencers uh with respect to uh sensoring of course information providing availability to the users and of course in case especially there was a really nice paper recently on arbitrum in case for example uh you have the the period where you can actually as a uh L2 L2 user try to uh disclaim that the transaction is possibly settled on L L1 um is actually relying on being able to communicate and not be censored to before you reach the L1. So for that there start to be some interesting challenges with respect to

that that actually are possibly to be solved uh using some form of external communication to the users and other part we definitely see that I think is being actively discussed with various frameworks is interoperability where we see that there are various approaches uh with respect to the challenges we foresee. So we see ourselves uh to look into the ways and actually how can we improve uh on the current state-of-the-art solution using our unit capabilities as they are. So in general uh what we want to provide is a defense in-depth uh mechanism to existing L2s. Uh so this is something that we definitely foresee that the future and the communication

capabilities for L2s will go in the multipro direction. I think there was already a thesis about it recently from Vitalik with respect to looking into always two out of three solutions. This is definitely something uh where we foresee that with our space guarantees we can actually provide a very comparable uh security guarantees or even better than what we see on earth. So this is definitely something where we foresee that we could play a role and of course in general this also helps with aspect of decentralization to the current systems as they are and I think what is maybe more interesting to the users itself is to provide additional uh preconfirmation. So instead of waiting for way too long time before the transactions get actually finalized on the L2 or the when the sequencer itself

is not able to communicate with the L1, we can try to provide additional guarantees which we call preconfirmations uh where we did envision to have a dedicated communication channel between the L2s and L1's where we can basically quickly validate the state update is is valid and later on try to provide this form of attestation to the users on earth which actually as it is aims to provide uh a good way that can be deployed across different L2s as well to provide additional security to the users itself. Here is the roughly outline how it would look like. So as a user you're interacting as you're used to with your favorite L2 and then the L2 itself uh would be communicating with us with our validator running in space that will provide just a quick state update of the

of the of the root of the market tree and just like provide an add station from within the space te in space provide this information to the L2 on earth to actually achieve uh this form of brief finality to the users before it gets settled um on on the L1 itself. So overall with this approach we aim to balance uh the aspect of security and usability to users while I think uh in general helping the ecosystem with wider adoption uh as it is. So for that uh we are currently actually in uh looking into the capabilities and the implementations of several L2s that are currently uh in our pipeline and we hope to roll it out uh soon either to our own

L2s but also uh future L2s as well. So summary what we can offer to L2S we covered I would say uh most interesting part that might be providing uh interesting capabilities with respect to the mixture of computation and communication resilience. We definitely want to provide uh step-by-step rollout to selected L2s. The reason why uh step by step is of course that we are rather a small and young team and we cannot integrate with with most of the L2s but definitely what we are trying to achieve is to also have standardiz solution that can later on enable uh other form of interaction with others and of course um as we are rather growing we have

announced recently our preede and are in the last process of finalizing our our seat. Uh we envision that this will be something interesting for us as we are building the infrastructure as well from currently having a single satellite to having having in the future a full constellation of satellites as they are. Uh we are currently in the last stretch of finalizing the white paper with respect to actually what the consensus how it should look like how it should scale moving forward finalizing the specs of the L1 and also as mentioned looking into the aspects of the integration and of course we are always open to discussions uh with respect to talking with various L1's L2 foundations uh looking into various providers that might be actually interested in maybe

considering a a pilot with respect to custo deployments or other type of solutions. In general, we envision that we are building a permissionless system where other peers can also later on uh run their proof of satellite uh in the future and deploying their own solutions in in the in the space. And actually what are we currently providing is the the last item which is we can think of it that we are already having a public preview of our capabilities using the toon number generator on earth where we are currently looking into applications regarding NFTTS of course but also look into solutions uh related to to AI um and other actually integration of randomness into various consensus protocols but also other cryptographic

protocols that heavily need random as well. So, uh I'll be here all the time. So, happy to uh to find out more. Uh get in touch as well. Uh we are actively working on it and we are currently uh building the community that is of course uh and should be space built and u happy to to discuss more. Thank you very much for your attention. Thanks a lot Philip for wonderful presentation. So anyone of you got any question? Perfect.

Uh yeah, I have actually three. Uh so the first one will be like if I'm not mistaken, uh all the teeth require some kind of key ceremony, right? Uh like correct. Yes. So uh what's your procedure for that? Sorry, can you elaborate? What do you mean by key ceremony? Uh like well they need like to be to in order to be trusted. they are not like uh you know like trusted from the start and it's like some kind of uh in in initial like you know like I'm I'm not sure key ceremony is the correct term for this. I see what you mean. Yeah. You mean like when it for example comes to uh let me go back to the overview of the various TE there.

Exactly. So for example when it comes to Intel TDX or AMD uh you are basically at the beginning uh communicating with Intel or others to basically basically get verified that this is a valid CPU that was issued by Intel and later on uh the attestation has the certificate chain from Intel that you later on uh you can basically trust and the secret is done by Intel so you do not touch like correct you have secret in the chip you cannot access it and it's like verifies that it it's like this exactly the part I think where ARMCA and especially I think The risk 5 is very attractive because you don't have to trust Intel to be generating the keys on the chip, but you can actually just get them generated on the fly basically. Uh which I think it's a very unique property that we we are aiming for. So

that way even if Intel claims they don't see still there is a bit of a trust issue, right? If they actually don't don't see it. Yeah. So this is something that actually the other technologies aim to provide and this is the reason why we are also actively looking into them because it's not only providing the better physical isolation where if you for example look at Intel or AMD they don't consider sophisticated hardware attacks as a part of their threat model but of course you still have to consider the aspect of the key generation process that we aim to uh resolve in a better way. Uh another question I uh sorry yeah uh thank you that's a very nice uh cool idea. Um my question is um so you know in a blockchain you are supposed to

run a decentralized system in which different entities run the nodes. So my question is here who is running the nodes that are in the space and also is isn't there a concern of you know when you give the node um to be launched into space that has to pass through a company that has a rocket and and during that time they can tamper with the hardware right um and and all of the nodes that are going to be in space have to go through that um you know company so is there any you know risk there or trust assumption that you have to be Yeah. So first uh the comment on who can run the note. So as we kind of foresee ourselves uh moving forward, we want to create a blueprint basically of a suitable design

where later on we can share with the community and motivate people to run it because of course we envision to have a regular incentives for running and participating in the network and the blueprint should be there for the sake of of course decreasing the cost so others can just easily go to a manufacturer of choice and get their satellite built and launch. As mentioned this is going to be as of now it's more expensive but we envision the future where it's going to get cheaper and of course there will be some upfront cost but overall we envisioned that our tokconomics will allow to cover the cost of you launching same for example when it comes to getting uh certain amount of stake here basically you have to get a certain amount of hardware to basically be able to participate in the network.

So this is on the first part. We are still finalizing the details but are happy to brainstorm on that part because it's something that we hope to build together as a community. And on the second part of course there are possibilities that u the malicious space provider let's say who wants to launch provider can temper with it. This is something that we have to consider but there are also solutions on how to overcome it. So for example, due to the aspects of launching and later on deploying the satellite in space, you can for example create your model of the satellite on how it's supposed to, for example, behave when it's properly sealed and not tempered with. And of course, if the launch provider, for example, was tempering with during the during the launch, certain aspects of the physical properties of the satellite

would be different. For example, the inertia and others could actually later on serve to detect such type of behavior. So this is something that we are looking into. There are already solutions. This is always a matter of cost and the the gains you are getting from it. Um I have a questionard regarding your L1. Are you planning to have any computational u capabilities on L1 in order to open to like space economy to actually allow to run any computation on the L1 or is purely for the consensus?

Uh so you're completely so I think I mentioned in the beginning so of course we we aim to be a smart contract platform meaning that actually we will have uh EVM capabilities or uh basically currently we are we are aiming together to have EVM capabilities so you can just deploy the smart contracts you're used to on our platform where we of course want to integrate it together with the T itself. So basically you will have a confidential uh virtual machine where you can deploy your stuff but run on L1 or L2 on on the L1 on the L1 on the L1 in space. Exactly. Yeah. So you will have of course the security guarantees for the consensus client itself but also for the virtual machine itself. Yeah.

Okay. Perfect. Thanks a lot uh for your answers and uh we are moving to the next topic. Thank you. Thank you very much.