ZKPodcast: TEEs, SGX and the Graphene project

Transcript

[Music] welcome to zero knowledge a podcast where we explore the latest in blockchain technology and the decentralized web the show is hosted by me Anna and me Fredrik [Music] in this episode we sit down with Lukesh from Golan and Cha che from Texas A&M to talk about trusted execution environments and the graphene project [Music] before we start we won't say thank you

to this week's sponsor app graph app graph is a project that's built in an extensive collection of open access research papers on cryptography distributed computing systems and blockchain the team behind a program see the high price put on the world's best research as a hindrance to scientific progress and so their platform is designed to break down barriers reward researchers and promote open science they start by offering unique content from technical conferences and course material from leading universities through an integration with their Orchid registry contributors can develop and establish a professional profile based on their entire body of work on a single platform if you're a researcher who would be interested in such a system find out more at welcome topography o or

check out the link in our show notes to a curated list of cryptocurrency course materials thanks again to epigraph and now our episode with the guys behind graphene so today we're gonna have a conversation about tes this has come up I think a couple times on the podcast but we've never really delved into it and today we have two guests who can really help us do that first we have Lukas who is the R&D lead for at Golem and we also have chai che who is an assistant professor at Texas A&M University and was the original researcher and author of the graphene paper which is something we're going to talk about as well hi guys hi hello Hey

and of course we have Frederick hello I think I mean that was a very brief intro I think it would be really great to hear a little bit more about what you guys are up to what you're working on and maybe a little bit about your background Lukesh why don't we start with you also I am a product owner of a graphene team in in Golan also my priority interest is researcher and entities and moving going into the innovative areas like tes which is which I think are very great and in future it will be a very popular area for for many others I'm a mathematician a PhD my interest is very

very theoretical but as a professional job I'm a developer for many to many years in goal we are starting to to create a decentralized marketplace when people can share power to others and other people can access the external power source of conditional power cool I think and as we continue our conversation we're gonna find out a little bit more why the work of Gollum sort of led you to experimenting with tease and wanting to work on it but I think first we should hear a little bit

from JJ can you quickly tell us who you are and how you got into this space right now I'm a assistant professor in Texas AM University I just start teaching this year I've been doing research in mostly in operating system area and security area so one habit of mine is basically I like to deep really deep into the system so if I look at application and then I could have solved the problem I go into deeper and deeper so that's why I really find things like hardware especially a security hardware like EE really fascinating because it was sort of solving a problem for the fundamentals right and so my history with two years so I we started working on graphing project and graphene turn

out to be a really good fit with EE so at the time there is sort of a conversation I hear we should start doing it if Co te and Intel most-read interested this technology and so I said ok why not why don't I just go there and do an internship and it turned out that's that's where all the we can start so I think before we dig any deeper into this episode we should start out by defining what a tea actually is so maybe first and foremost what does te stand for I think tea in general would stem for trusted execution environment it's usually describing a hardware that's designed for bootstrapping yours your trust so when you put a application on a

platform you don't actually trust that platform you want something to help you start with that trust and that is the Hauer and it actually has a very long history is start with things like trust processing model TPM and and it also started with other kind of environment for example arm has their things called trust zone which is basically environment that you isolates your application away from other application and also trust operating system basically the whole technology has grows until we hit a point we have things like we draw to not trust anything beside your application in your hardware which is support the birth of SGX and there's a lot of few technologies coming out after as chancers including arm AMD they

all have their own version of it there's also the open-source version from for bliss 5 right now and basically this is a trend that's going on what would the use case be for like you mentioned these early projects like TPMS like what were they used for and and what are the use cases that have like demanded that this scales up so I usually I was separating 2 to 2 domain one is on the desktop or on a personal computer or always say in decline it right usually you want to start some execution that's really sensitive for example uh one very common example is the digital rights management let's say you have some kind of video that is copyrighted you want to ship it

to users machine cause machine you want the encrypted I saying there you have this key that you want to protect right then you can use these either TPM or SGX to make sure this Enclave was shifted to a platform that's secure that would not leak your secret that's an amazing example you just mentioned SG a lot we haven't really defined that we've defined T's so SGX that's a type of T so X is definitely one type of te it's basically introducing a technology like o Hardware on clay so the idea on clay is if you heard this word on clay is basically usually we say a culture on clay is a area where people live in and they'll isolate from the rest of the

world so the same thing as hard on clay va's you run some program and you put on your computer and that was completely isolated from the rest of the computer so you don't have to worry about operating system being compromised or other i/o device remember it's actually very strong guarantees of security so as chips is this step will suffer or extension is an extension that cabal recent Intel CPU if you have like AI Phii said the CPU is likely you already have it you should check it out you probably have it and if you turn it on and you can use this technology to protect your application and this is an Intel this comes out of the Intel research group where does it come out of with an Intel it actually has a very long history internally I believe they have to move this thing for a ten year

already so they have some very long time to make sure this is as good it's correct it's secure and they put it out there and this is now your product line right now you can actually get them is off the shelf there's still some see whew that's a half at the efforts of the I Phi I'm sorry II five the more server class if you doesn't have it yet they probably will have in the future I think one T or are particularly the term Enclave is something most people have heard about in the context of Apple iPhones so iPhones all have enclaves or the newer models and they use them to store like fingerprints and your face ID sensory data and all that kind of stuff

so it isn't reachable from the rest of the operating system you can't install an app that steals your fingerprint stuff like this but that's very different like they have a specific chip that is this Enclave and it's not connected to the CPU how does that compare like is into cutting these two things together or is there some benefit like additional benefit to having the Enclave connected to the CPU I think we clarified that the Enclave is a software and SGX and key rather refers to hardware's difference is tremendous right because such chips like in Apple gives you only one functionality and lives as a

software as a part of application that can be run on your computer gives you opportunity to write your custom applications the benefits are absolutely unlimited I mean almost everything you can do with your own application the values that this application brings your customers can be achieved also with enclaves and with HX right so this the difference is it's like two different worlds it's a specific purpose yeah so I I partially do agree with Lucas FL you know there is definitely a different use cases and different ways of application interfaces how do you programming them I think there is

actually a deeper security argument in tariffs do you have everything inside one processor once if you or you have these coprocessor models that you have a separate chip or a separate processor for somebody in the arm trust Amato they actually have another CPU was handling it and that was deeply in related with the kind of problem where you want a better performance you want to have regular program like like the program you run every day to run in this Enclave or you want to basically spend effort of reprogramming everything and have everything just running on this little chip and that is there is this argument there for the trade-off between security and usability from what I understand the like would you still call what is on Apple on these Apple phones

as it's still an enclave is that still a tea or is that something else that's a tea I it would be yes but it's like a simple one I guess it's like it has one purpose or it's like only confessional purpose one yeah that's right as the SGX the Intel SGX from what I understand is you can actually run some of your application within it or would you run all of your application within this well that is actually really interesting questions I also can't debate that we always go in with right now in the attachme domain is do you want to take the whole application in there or you want to just rewrite part of your application or you want to partition whatever application that depends on what is just a 3d model do you want to reduce your risk of your

program being being attacked because things like heee they are not silver bullet they don't solve all the problems right if you have about a vulnerable software you can still have problems security issues okay so you do want to consider which one you want to trust and you want which part of you want point into this te environment what does it actually protect against like is the idea here that if you had a bug or something running on your CPU that was malicious it couldn't affect that is it just to save it from other things that CPUs can do so usually was protecting us against one very strong attacker as the host attacker so imagine your operating system is compromised or even worse your bias is compromised where even the whole

hardware is compromised like yes I am the MIS traitor I have SS to the machine physically and I can change a lot of thing for them tol if we talk me the extremely they are a tag light for someone Rho hammer attack or something like I can steal data directly from the i/o buses or your memory buses SG exes these things could actually protect you from for all these kind of attack potentially because they encrypt everything incredible data when they are sending out of the CPU as a host can I even access my own SGX it is sort of like it even protects it from me you're not supposed to you're supposed to see just encrypted data on D realm and you are not supposed to see anything

out of it to answer this question to solve this problem I think we need to better understand what SGX gives to the users what is the value of T is in general and at the beginning to give a very good example I will give another example consider this in more in the most countries the law prohibits to send sensitive medical data to cloud provider and for his data on clouds and make any computations transformational deviations of this data in external cloud this is

made for very concrete reasons imagine this that cloud providers enables HDX on various machines and what some company or Institute State Institute do is take this medical data encrypt it and send it to the cloud provider to the cloud and store it in the cloud in the encrypted form the car provided any administrator cannot access this data but the this Institute or any company can perform any computation or matific modification remotely on this data right so who is

been a who benefits from this solution the cloud provider does not have any direct benefits from this the end users who are accessing services provided by cloud are benefit us right the Dalek benefit or does not use HDX directly in this use case particle but he is using external HDX so what is in the in the farthest benefit for an user it is ability to access external power computational power and storage we found floating confidentiality and privacy right and

that is that is the benefit so if Anna is asking what do I have when I enable lgx on my computer it's pretty complicated because in this use case but I have described the benefit or is external user right if the s3x just the place where the encryption happens yeah I think that should be quite obvious but in contradiction to this chip installed on maybe I'm wrong I don't know details SGX has maybe not dire ability but in some form it has ability to connect to

external users so it gives a lot of possibilities opportunities it enables lots of use cases and this is this is the power of cheese and as Jax I want to dig into one final question just song kind of clarifying how SDX or or any other tea actually works and so you already mentioned there so you have encrypted data on RAM on disk whatever send this to the tea and you get something encrypted data back out but for this to be trusted there also has to be some proof or some verification that this computation has been done correctly what does that output look like what does it does the proof look like that

comes out of this thing your question is it's a very good question it's a fundamental question and a lot of people ask this question the X chips has sealed private keys inside hardware and they are used to secure communication with this GX the most difficult step is how can external user the end user or even a co-located application that is going to employ in place know that it is speaking with an end life how can it know this speaking wave trusted

component it boils down to attestation process and the moment is solved this way that Intel's provides services that can confirm that you are speaking with Hardware real as G X it is it is very high level and have you not getting into the details so is it sort of like almost like a certificate authority thing that Intel knows what keys it has on its processors and and can like you have to reach out to Intel to say that is like they attests to the correct results or this work so I'll extend our well Lucas was saying you know it's a gritty first

always really up to implementations you know SGX is where other tech te will have different implementation but it eventually I'll boil it down is these platform need to keep your identity easy to prove that this is a real horror this is a good horror this is not a highway that was simulated or emulated by the attackers right that is fundamental that you want to provide right and so place of quote for Intel what they do is they have this master key or master sequence was priced into it inside your CPU and because they have had the hold of all these secrets that can use that to verify that you were power is correct some other platform might use different things they want to do for example TPM I use more close to what the certificate

system you have been using but although mostly they are using these crypto tech thing like the the public key structures and those are thing to make sure is your your platform is secure but by by this definition what you just said it sounds like you have to really trust Intel absolutely like are they publishing anything are they are they sharing any of this publicly where you can actually verify it yourself or is it like a trade secret they they used to be that you do have to ask Intel all the time to to verify everything recently they have become more open on this so they have do something called these TCAP which is allows vendors for example that say you have a cloud vendors you could absolutely build your own attestation server to verifies the secrets on their

own so you don't have to go always goes back to Intel's and do all these things right it do sort of moving unfortunately is it they move really slowly but they do sort of move forward instead of making it more public and will open because I think yeah if that isn't shared I mean at least our community we always hear this it's like the reason we prefer open sources you can check it you can see what it is you can run it yourself you can make sure that it's running correctly when you're dealing with sort of a black box like trust us we have all the keys we keep them really safe I promise it's hard I don't know and I guess this is their response in a way their their attempt to like start opening it you sort of just mentioned there that

you work it's like their work with vendors or their work with other I guess software setups where it's it's they're not it's not like an individual would interact with an SGX ever but like there are teams or groups that want to use these setups for some purpose and they work with until I guess to do it so they have this back and forth they have this ability to do an attestation that's right if this where this this is kind of where graphene comes into right in this space oh yeah that's right so remember I mentioned about the desktop usage so using a CSS on my clients machine your usage machine right there's another model or another use cases is basically

using it as a cloud and the idea this is basically you have some application you want to resell Clow do you want to trust a whole cloud completely we just want to trust one machine or one CPU inside a cloud and so if you can basically reduce the trust of the inside color that would be much better so there's a lot of cloud vendors are really interested in this direction helping down Amazon I haven't heard of Amazon do anything about it but like for example Microsoft address or IBM or Alibaba these call vendors are starting to get into building their own Oh so Google has their own version of T as well I have to say that they were actually basically putting together these technology and

said okay I would want to have this kind of our local environment to protect your applications right and so that that just sort of had it had kind of give a space of frame were like graphene because well to be honest user a lot of time we're lazy right they want don't want to poor application into this is platform and what the role graphing is basically very simple you have application running else around clay you can use this framework to move inside on clay and run it as it is and we're basically solving this trying to solve this security problem and competitive problem or vacation what was it like before like what maybe describe a little bit like what is this set up now or before now and how does this change cuz like it sounds like are

you sort of an enabler you're sort of like allowing that movement easier yeah so idea of SGX it used to be the cases where you have to reprogram everything right you use well not this room we put that every line of a code but you do have to take a part of code you have to rethink about the security problem because you have to build the interface between unclaimed and not unclear code and you basically have to rebuild basically take your system and say all right I'm gonna take this part secure secure it and put in some clay so there is this difficulty of implementing any as shields program I think if you a really good example in Berkeley I've got like undergrad students come to me and complain to me saying oh I'm we're trying to get as

you're working but it's taken us like a month or something just to figure out how to do this thing it's really complicated and partially graphing is really saying no you don't have to do these things we do it for you but you just take a program and we got to move it into Enclave and also we want to make sure it's secure and that's actually very important thing if you had now security is not compatible at all why was Gollum like what was golems role in this like God's a golem as you described before you guys want to do kind of computation and actually from what I've always understood Gollum doing is sort of like marketplace for off chain computation so where's what's the relationship here a very direct relation to graphene and to SGX and keys in

general because it is for many users for many potential customers of your network will be willing to use his network as our consumers are not willing to share the data input or output data like a Ledge video company movie company would like to share buy a new camus v any any fragments and pieces of this new movie to anyone because it is very apart so we recognize where this it is a very huge market for confidential consultations and it was very natural for us to think

about employing SGX but a structure said it's very challenging to do right from the scratch and you as GX application or an active maybe as UX applications not a very very good phrase to employ as Jegs in your epic application right we needed that tool to grab an application and around it and around it in a Jax just just acted and it's turned out what the graphing is the way we should go so we get involved in this and since been really we are very happily cooperating with church and on and in improving the

graphing and getting back to the beginning it is the answer for the confidential itical and huge step towards very demanding customers I have two questions actually one is related to graphene the other is kind of going back a little bit to T's but first on the graphene thing are you explicitly trying to target more than SGX or is it like an SGX specific solution so I would say you know in general we always trying to target more platforms so fundamentally graphing is supposed to be very thin layer that trying to saw compatibility problem insecure isolation problem or application on as many puffle as

possible for example a lot of time you run into a problem what if I want to run a Linux application or Windows or Windows applications in it and you have this interface compatibility problem and the way to solve it it usually is you feel very thin later to translate these interfaces from one to another and graphing is supposed to be this ultimates interface or translator a translation layer that's reconcile everything but of course it hasn't it's not there yet but our goal is basically make it easy to pour to everything and s shiksas has always been our what about gold go to an example because it was really how do I put it exotic it's not the same as anything other platform

you've ever seen and many because now you take this interfaces and you foot the sacrimoni upside down you used to trust operating system now you don't trust it anymore how do you solve how do you even you see system 8 yeah how do you open a file how do you even okay some memory not become a very big issues and you need to make sure all these interfaces are secure and so that's why we kind of believe if we can solve the problem as chips is you probably would solve the problem of a lot other powerful as well so we're basically using that an example of a push our research goal and who knows maybe eventually we would we do need the layers to solve a new hardware you know there's always going to be new how a platform that coming down with you offered some copper coming out but if we

actually need something to run some new some Oh what application you want it well graphene Oh we'll come to rescue and help you solve this problem did a golem look at other platforms before deciding or did you just know like STX is the furthest long this is a watery what we want to shoot for from the Gecko as such a say said I agree with him but it will be great to see some competitions in this it is very hard for everyone as usual competition is very healthy and I think the graphene is able to adopt different solutions for different keys and this is the way how it should be but the various another view of this because we do not have at

the hand ready to use our solutions for tes a bourbon eject and there are brass but not with this scale and at the moment the scale matters right so it is absolutely natural but our interest is in Intel's edge ax and we are thinking very very strong about it about getting back to the beginning it would be nice and I hope it will happen but our phase will appear and they will be for our use and the end the graph and I think we'll be ready to adopt it as far as you know there are definitely a lot of TV was

coming out for sure so a arm has up so MT has their own version it's called it's called se ve arm I believe that they are properly I'm not sure what is on market yet but they already are building this thing called crypto Island which is basically trust zone plus our encrypting facility that help you solve this current problem and so they all doing this guy thing what they are saying is I do think they have become cautious right now because they want to make sure they're doing things exactly rights do not have more ability as we can see there's they were a lot of fun ability on this kind of platform and it usually make people doubt the security of this plat for a little bit and that's why they're being

been careful about right now so we will see but it kind of pop cards in a while we want to get back to that topic a little bit but before we do I wanted to ask a sort of general question on teas that I it popped into my head as you were talking my use cases let's say you have this use case where you're running in the cloud you want to upload some data transfer that to a tea decrypt it run your computation and then encrypt it again and get it back out so if I'm decrypting inside the SGX I have to supply my decryption key to the SGX environment and how do I do that if I don't have direct access to the

machine how do I transfer that over the wire like how does it go through Amazon's Center or whoever Center and getting to this react without revealing my key it is solved in a different direction opposite direction the SGX is generating the pair of keys symmetric keys to communicate and it is very easy to give you to the end-user the public key so any any data that you encrypt with this public key can be only deciphered by by this ank life all you need to know is that this public key from comes from Rio Rio as yet right

this is the part of actually at the station process right yeah and Varys and Varys and infestations involving this sea so that makes sense so essentially I asked the SGX environment hey please generate me a key pair you get the public key back with along with an attestation from Intel saying yes this is a key actually generated by an su X environment I encrypt my data with that public key upload it and then it runs through the ACS GX how do i decrypt it when I get it back it is much much much much much more complicated than

that all right I will try this way first thing that we can do after receiving the public key communication key to the anklet all I do I am generating a symmetric key for a phishing encryption I encrypt this symmetric key and encrypt with public key communication public key generated by an anklet and I send this encrypted key to the enclaves only and clave is able to decode it so at the end I and

the enclaves holds the symmetric key that is used for fr4 communication and in the next steps in the formal communication this semantics is used for communication right and it is very simplified process but instead of yeah what you're saying is instead of just me encrypting the data with a key I God I encrypt the data and the key that I want to decrypt with and send all of that over and then when I get it all back I can I still have that key and so I can decrypt it because the ss6 encrypted it with that key that I gave it that makes sense thank you this process actually called secure provisioning and we usually said

Ducati is pretty much useless if you don't do provisioning because you do need to move some secret data from the remote machine your local machine to process them otherwise it doesn't really make sense to run any cutters and so of course these kind of provisioning the shipping data needs to be on a secure channel and what Lucas is just describing is basically how they build a future at all is actually not too much different from things like HTTP that you build like a TLS or SSL connection it's the same thing instead that you do the key exchange you do you use either secret key / private key to encrypt your data and then you send it it's pretty much that if there any in that sort of encryption do encryption like what you're just describing are there any

other tools that are being used maybe tools we've already mentioned things like filming fully homomorphic encryption or NPCs or like is there anything else right that's an interesting question so research perspective there's a lot of people actually you sort of saying you know it is trying to use heart uncle to replace homomorphic encryption or replace multi-party computing i did is basically because you have this environment that everybody could all trust you don't have to do this step like extra crypto step to make sure your data is secure you can just move the data to the platform decrypt data and process them so actually making a lot easier make these things well it's actually not secure of these things of course but it sort of

replace it and make it much faster much more efficient that actually leads to another question which is exactly that like te es are often compared to fully homomorphic encryption or even like zk snark kind of check-in points and I've actually I remember in an episode we did on NPCs we mentioned this and T's seem so different and yet they're often lumped together when we put the question I put a question out on Twitter sort of about T ease and I definitely got some feedback from people maybe half of them were like oh this is a interesting cool thing and then half of them were like you cannot trust them why would you even talk about it so how do they hold up to the fhe MVC or like the ZK snark paradigm yeah I

mean I I would like to hear if there's other different perspective on the industry for sure but so far the Academy was stupid we usually said there's two different approach right there were the math mathematical approach basically you solve everything Sorey problem by math like using something that it can prove you can verify in the awkward it's awkward on level and you can look at principal hip thank PC and you can verify offices is secure another way is implement is secured by implementation right you have or end by engineering you have a hardware or some kind of operating system why provides a label or whatever you do and you trust implementation that and that bootstrap your security and

I'll say it's not really about which ones most secure is more like they have different challenges like so your secured by crypto all you've got the challenges of you know efficiency and whether they might be attack at the algorithm the whole when you look is CPU what SGX is they have a heart attack software attack these kind of things where you have to always solve is another and so it's all about you know how would you you want choose do you want to pick what do you trust they could make you say fundamentally do you want trust CPU vendor doing their job properly and design this hour so people and so that's sort of the trade-off right it's sort of like do you trust the group who have created the hardware to not have a backdoor or do you trust the

math that's exactly the kind of the circuit judge Nero soffit right now and it's actually not complete self as you can see and so or look into these kind of problem let's see yeah you mentioned that there has been bone your vulnerabilities discovered in STX what kind of vulnerabilities are those and how have they been patched how are they and like how does the platform recover from those so I think first of all there's always the sulfur them old building and these are not new right it's the same kind of motor ability you see in exhaust surfer or desktop application for supper like you know hot breathing bugs you know it opens it so you know the kind of same thing that happens in unclear as well and especially you have very strong

attackers who have control of the operating system on the other hand at hard level we also sees things like site general problems where essentially you're sharing your CPU core with process or water or other operating system now they could see a lot of information being leaked through your share resources and based on that well now we also have Howard vulnerability that based on either speculated executions or memory protection by passing things I expect her and mailed out box there's actually one that come out last year or we call for shadow box is basically the meltdown will kind of type type of vulnerability that kind of pipe to SGX and this is kind of a big things when you see this thing day they

potentially could leak these master secrets that being generated by Intel CPU that verify your platform and as you can see if you leak that secret then you're gone the security is basically ruined and so you do have to either solve this problem or pass your horror to solve this problem and that is that is for treating the reality of this technology yeah I mean so with a lot of the normal inte old bugs that we've heard about there are various fixes of like compile it with this flag to turn off that feature or like the latest one I think they say turn off hyper-threading and you don't have this problem which is I mean yeah you don't have much choice but

hopefully they fix that in the next generation then you can sort of work around it even if you have to wait for the next-generation CPUs to have it actually fixed it's an interesting thing to think about like what could go wrong both on a social level like what happens if Intel leaks data on their end so that anyone can attest to anything or what happens on the hardware level something I'm super curious about is like what what how temper proof artis because you you have these like super involved attacks being described by people when you have hardware access like you dump your whole computer in liquid nitrogen and then you you freeze the RAM and then

you can read off the RAM you can like detach it from your computer read off what the RAM had post facts like circumvent all of these like shutdown barrier tricks really you have cool exotic attacks like that like what happens with those kinds of attacks on a TV environment what if I like D lid my CPU and look at it with a you know electron scanning microscope or whatever so I think in general the assumption is is really difficult is I want to say completely impossible but it's almost close to impossible that you all could open a CPU and look into it because everything is at the scale of nanometers and you pretty much couldn't really see anything and that is also kind of moving

the direction of DRAM you know you actually couldn't really open DRAM and look into it if it has been really is down to nanometer level but I've heard a different story as well like some people could say some CPU potentially could throw a hold on it and look into it probably not ease Intel CPU but more like the other kind of CPU could actually do that but in general I'll believe is hoping the CPU packages there's really two secure and and cannot be tempered with but you're right this is interesting questions and where did the technology I believe is even if people could do that it's probably really expensive and so it's kind of questioning what is worth it to actually do that that's kind of amazing that idea and so I think one of the critiques

seems to be like you just described this sort of human factor there social factor but I guess the big question is like are any of the other paradigms ready like if if math if crypto like cryptography and cryptographic concepts can do the same thing or can you know securely do the same thing why would you even use a tea why would you even use SGX and I'm curious what you guys think cuz I'm sure you're coming into contact with that like do you feel like T's and s GX is the thing you do right now because because the math isn't there because the cryptographic stuff isn't there this makes sense today or is this like actually this is going to be something that will be a long-standing solution

and is useful for specific use cases whereas the other stuff could be useful for something else at the moment for today I get in touch with physical snark recently and I have some thoughts about it I listened to a solution to my approach using zero knowledge proof and I think it's very great approach it's a lot of maths you don't have to trust anyone it's tremendous great and I think it is a very good approach when you have a very little data but are very sensitive like transactions it's a tiny data with a big value and

then you can use zero note proofs for such a from such data from my point of view when I have a gigabytes of data to confirm to make computation and to output gigabytes of data using zero no truth as it is today is no go for it so performance is a key indicator for for me picking us as a go on right because at the end what matters at the end and the end person end user who wants a service with a particle or our factors like like privacy enabled like confidentiality of his data and the

answer is yes for it I hope that some future zero knowledge proof will have enough speed to perform some computations to be also used as a service for remote conditions but for today I think due to a processing very the data with a huge value that's their answer the question right that it there are certain use cases where even today these other paradigms NPCs FA cheese snark Starks would be more useful but in what you've described here is that if you're doing like massive computation they're just not fast enough just speed wise they're in in such different worlds but you know the process and the development the research it progresses

pretty quickly on the crypto side but even then you're dealing with a separate problem which is what graphing is trying to address right like how do you take an existing program and run that there is no one even trying to think of that in the snark or fhe world where it's just not on the table at the moment like it doesn't exist whereas taking an existing program I mean in golems case if you want to like the purpose of the platform is to sell general-purpose computation it's not to sell you know I have to spend six months writing my snark circuits before being able to do anything so even if snarks were fast to

compute like having that lead time to write your program as a snork circuits is already the showstopper right you want to be able to take whatever our blender application and render or something then you need to take that software off the shelf and be able to run it and so I guess with graphene that becomes a possibility and we're in the crypto world it's not even you know considered at the moment so you know maybe I in a hundred years we'll be we'll see I guess yeah that's interesting yeah I think that really helps to kind of pinpoint what graphene is doing - I have another question about graphene in general which was like you we are using the Gollum use case here but like is it for all applications is it crypto specific is it like what's the

I'm imagining it's more SGX specific I guess your questions would be a bell you know what's kind of vacation you want to run and what kind of platform you want to run the top on top of it right as chances is one of the preferably one run but we're talking we want to run different applications different more different platform and more different more different different kind of hallways yeah different tea is basically exactly different te would be a really good I will hope in the future we're running on arm or run on AMD they'll be really awesome a photo application I think you know it's hard to make every application run for sure and unfortunately try to be general-purpose but it's really hard to have all the plication running right crypto of course

is different there won't any kind of like computation computation intensive workload it would be really good example of course there are things that are more Isle intensive or more Moore's administration kind of types things maybe we don't really have to run them on top of it for example you don't want to run configuration panels inside uncle you don't have to see that there's a water because you don't want to run inside I assume there's like some relatively major restrictions like you can't access the network and things like that yeah I mean you could you could actually use SSI oh but you do have to add protection to it usually we we call this behavior what we call she'll she'll because if

these building some kind of protection to it it's end to impotence you know what payloads or into a protect your your disk I say yeah my only last questions are a little bit about this like backdoor and a lot of the cryptographers that we've spoken to would also that seems to be the one thing that very much worries them the human factor this idea of like somewhere there could be political pressure or some pressure put on a company or an organization to add a backdoor that nobody knows about and I wonder what you think of that like does that is that something that you is it a legitimate concern or is it overblown in your opinion you are comparing two groups were it's a mathematical proof of

that you can read the data right because there is no back doors and verse and an opportunity for a manufacturer of the edge X or Alberti's to provide some back doors or Falls or anything like that for number of reasons right politically those internal politics reasons and control or anything like this as far as I understand your question right your question boils down to why do you trust Intel and the answer is I don't think that we are trusts inter very very much

I hope that the interest is very mature and responsible company and they can predict the providing any bangles or anything like like this in my products but looking for the future right maybe there will be more teas on market and the competition will change the whole the whole technology how we look at it how we employ it how we are using it what are our expectations to this to this technologist right as you said in this is human factor plays very very strong crawl and of course many thing can

happen if it is happening right now or if it will happen in the future actually I don't know I don't know oh I know we are developing very very good technology despite any flaws of hardware solutions right and we are quite good solution for users for developers for companies who would like to develop applications and provide services based on the confidentiality privacy and anything connected with is due back home and end-users I think that this our our position for for this this this aspect

of the objects and das in general what do you think trechie I guess I will kind of sort of say it's kind of ok ok this is my reason um so so that I said earlier you know if you are doing a different approach you've basically fate different challenges so if you are going with a approach of securing by engineering you're basically facing why you need to make sure everything was done properly from for design to manufacturing to shipping to local supply chain problem to to selling them to people installing them everything you wanted it to be probably secure and of course there's gonna be there's gonna be delta bail is if we can actually make sure all these steps are taken properly

yes they are these kind of issue for example last year and is the the bloomer reports about this big hack about their chipping ingeras of course it's I don't think there's a hard evidence of that yet but you know there is technically this is something that's feasible right so of course for the perspective we do want to have this process be open we want to make it transparent to see you know how how are these being manufactured how are they being shipped where they get we'll have they come from like this we want to know these information before we actually run any secret on top of it but in general I would say maybe actually if my perspective that I think for a lot of either software vendor or

service vendor it's acceptable because you know économique from the company perspective they could understand like how to use the killer a manufacturing process how do you secure shipping process how to secure secure the supply chain problem it might be actually easier to understand then the way that crypto work or zero knowledge proof work or a Multi multi party computing work you know so somehow is necessary of course there is the risk of it but this is the risk it can control they can understand and they cuts minimize me and so somehow that's why this turned out to be a more optimal so ideal approach for the industry that's what I believe that

I feel like there's two things to think about here there are like two arguments to make err one is Intel has more at stake than I do absolutely which is true for most cases like if Intel gets corrupted and things up they will lose more than I will and so it's okay and I think that's a fine argument actually up to a certain limit like at some points like if we're talking about the entire Bitcoin blockchain and 50 years from now then maybe that isn't the case but who knows the other argument is we can actually make this better so SGX is where it is right now you have to trust Intel but if they open up more or if a competitor comes onto the market

is more open there's things to mitigate that risk right yeah it is a very good point because if we how many users is using SGX and t is right now if using it directly or indirectly we should say that we are on a very early stage right and I said something about the open source and I liked it and I remember it because open source comparing to a closed source is is very well tested because it is has lots of users and the code is open for reviews and then GX for us very few users some developers but I think the T is in general are are the future for

mining services for many applications and with the time it will be more and more solutions better and better tested and the whole landscape of G's will change maybe they'll be more and more open - yes I just want to say thank you so much for coming on the show and going kind of covering this landscape with us and sharing kind of how you guys got into it why you're doing what you're doing and the ins and outs of the te world yeah this is really really informative so thank you yeah thank you for having us and to our listeners thanks for listening thanks for listening [Music]