David Sneider I Decentralized Secrets for Identities Networks & Platforms I Pragma San Fran 2024

[Music] for this next talk I want to bring on David from lit protocol to tell you about how we can rethink managing secrets on chain so it's going to be a lot of cool things here please welcome David from lit all right thank you cartic hello everybody uh as you can see my name is David and today we're going to cover four things basically the broad problem space uh we'll get into that in a second the solution we've invented and I've had live since February of this year that's done over 20 million requests some examples of various implementations and then the opportunity for the future so to start with the problem space this is something that everybody knows whether you're a developer a product manager technical or not and the fragmented nature of our world from a technological perspective isn't just something that technologists know but every single every single individual knows this who's a general user so it begs the question what's the common thread what's tying every system together because we know the world that we want which is to have agents applications protocols that seamlessly interoperate across every single touch Point platform Network uh for the benefit of businesses and individuals now it's not a secret but it is in fact secrets are the unifying Factor here whether it's a private key on a blockchain password on your social media account uh uba key an API key to access a B2B service and this is the specific problem that we have been working on for the past three and a half years at lip protocol which is a programmable Secrets Network and so the big idea here is if you write applications and protocols with functions that dictate where secrets are used across these systems across these networks and platforms you create a new class of kind of unlimited composability across all networks so we'll go into the architecture because this is a non-trivial problem to go after this you know no pun intended but key problem of secret management oh yes and we'll get into some of these these use cases but as mentioned the system has been live in mainnet beta since February of this year and processed 20 million requests to date so as you can see uh we've got three Graphics here and this is moving in order of scale from left to right a node what the network looks like and then how this scales and so let's talk about the four components over there on one of these single lit nodes and again the intent here is for a program rable secret management Network where the secret is not managed by a single party but is a decentralized distributed and safe and secure secret so first you'll see that the uh outline is a secure Hardware this is encrypted VMS uh specifically AMD which is their SEC secure encrypted virtualization this is a type of te we've written a specific operating system for this te to seal and lock this down essentially turning it into a black box that the operator of this box cannot get into because the secret material in there is the red icon the key share uh which is part of a threshold key which we'll talk about in just a second and then there's also an execution module and a networking module for data input and output so moving to the right up to the network scale this is a threshold Network the specific methodology here is called multi-party computation threshold Secret schemes the important word there is threshold this is how secrets are managed in a distributed context each of those key shares represents a part of the total key and in order to generate some operation like a decryption or a signature from the underlying distributed key you need in the case of lip protocol 2third of the network to sign off on this and signing off is not something that there's like a human in the loop this is done in an automated way we'll talk about a a couple specific examples in a second and basically at this network phase is where we are today we have a handful of networks live processing all of these requests but where we see the world going is towards a key mesh uh with many many networks not too dissimilar to the app chain hypothesis in in ethereum and Cosmos but in the case of of secret management and so just to talk a little bit more about how this is used to interconnect systems literally anything can trigger one of these functions uh where the output let me say this another way you can basically take an input from any system whether it's a user generated uh input whether it's an event that happens on chain whether it's an event that happens in an API to start this process of a request and then for a given request those three components are there again which is Computing inside of the te so it's blind creating a decryption or a signature from the shared key or fetching data from a blockchain or a user or an API so functionally you can think about this as a new primitive uh folks in the past have kind of compared the Insight here to ethereum like basically vitalik's big idea is that setting up a blockchain is a really big task now if you can make a blockchain programmable with smart contracts nobody has to set up their own node Network because you can just program the chain with smart contracts there's a similar type of mindset Happening Here but rather than having to go set up your own distributed Key Management Network you can actually just Leverage The distributed key and write applications on top of it and then of course you can see the outputs in turn can also go to any other system and these things can Loop over and over and over and you can create really sophisticated and and interesting protocols and we'll look at a couple um examples right now so this is a little bit related to what we just heard from Jason its story uh this is a project called verify it is uh an IP licensing network from the blockchain creative Labs uh inside of the fox Corporation and functionally the way that it works they have a polygon cdk blockchain and media organizations like Fox and time what they're doing it with their articles is they are encrypting that article go back here to the distrib Ed key because the public key is public and then setting a rule around who can decrypt that or view that asset and so the way that it works specifically is let's say fox writes this article it gets encrypted to the public key it's stored on ipfs encrypted so nobody can read it then somebody purchases that license on chain let's say that's Bob or an AI company or another media organization that wants the syndicate or reshare that content Bob what he's going to do is he signs a message and he does a one to many operation to all of those lit nodes that we were looking at earlier each of those nodes then independently check the chain does Bob own the license to be able to decrypt this content if yes each of those nodes use its key share to generate something called a signing share which you could think of as an authorization and then there's a many toone operation that goes back to Bob and so Bob collects those authorization client side and now he has the decryption signature as well as the cipher text and can go ahead and decrypt that content client side this is a process called identity-based encryption you could think about it as conditional or rules-based encryption historically identity based encryption was performed with a centralized Authority who dictates the rules and Provisions decryption keys but in this case it's being done with this threshold Network running inside of secure hardware and as you can see over on the right side of the screen I've just included some of the systems that are interl um where the chain is managing the license uh in this case they're using ipfs to store the cipher text and then their system also has implemented things like Apple and Google for uh being able to allow users to sign on and sign into this network without having to manage uh seed phrase because the key material for the user is managed by lit now let's take a look at another example this is a product called nebula um they're still in development but what they are building is a essentially an onchain life insurance company where somebody can go ahead and get some life insurance from a wide variety of providers and what's very interesting about what they're doing you can see I've got on the right side the image representing the user the user Alice goes ahead and inputs all of the requisites data required to get life insurance what are pre-existing conditions is she a smoker what's her age what's her location the standard form that would be required uh to get life insurance now instead of sending that private personal identifiable information to all of the agents what the nebula team is using lit for is that information is encrypted and stored on ipfs and then to determine the risk score we can kind of think about it simply in like a one out of 10 risk score in terms of what should the premium for Alice be that content is decrypted inside of lit inside of the secure Hardware environment then they have a risk score model that iterates over Alice's data 100,000 10,000 times and the output is that one out of 10 score around how risky is it to ensure Alice in this case and now the agents that are underwriting the life insurance can give Alice a quote qu without knowing her pre-existing conditions without knowing if she's a smoker things of that nature that private information remains private to Alice this of course gets really interesting Healthcare I think is the final boss of of web 3 in many ways um and so this notion of being able to compute over personal identifiable information in a privacy preserving way starts to get really exciting and the last example uh coming back to the crypto world this is an application called genius terminal it is they like to say that they're a sex on chain um it's basically a fully chain abstracted application and so you can see what I'm doing here is let's see oh I don't think the full GIF loaded unfortunately but basically uh the yeah sorry about that uh so the swap that I wanted to show is moving from usdc on on Bas into a into a meme coin on salana and unfortunately I don't have the ability to show it here but the swap took two seconds and the functional architecture that they have here is there's actually four programmable wallets involved in making that swap happen in less than two seconds there's a user wallet on base there's a protocol wallet on base there's a protocol wallet on salana and a user wallet on salana and the user wallets are connected to the same authentication in this case sign in with apple and so the transaction gets simulated oh I need to take my usdc on Bas and it needs to go to the orchestrator wallet on base then the orchestrator wallet on salana needs to hit a local Dex to buy that mcoin and then send it to the user wallet on salana so that transaction gets simulated the user clicks Swap and all of those actions happen in parallel so this is basically equivalent to the type of experience that somebody would have on a centralized exchange although of course without things like like kyc or trusting a centralized custodian um if you want to check this app out they're doing their main launch in a couple weeks but you can go to genius terminal.com slome and use the code get lit uh if you want to check out their beta uh that's all capitalized and that's where things are today now to just talk a little bit about the opportunity ahead these are two Graphics that you may have seen in the past the one on the right is from uh situational awareness a really nice overview around where are we in the state of AI or excuse me the one on the left and the one on the right is from vitalic post about his version of defensive techno optimism so we'll just talk about these two graphics for a moment to set the context for for what's next what's happening there on the left is uh counting the order of magnitudes of improvement in llm models and the prediction from this particular paper is that in about two to two and a half years basically the work that is done by AI researchers and AI implementers will be good enough uh excuse me will be able to be done by the AIS themselves this is like the AGI moment that many people are talking about that we're heading really we're heading like pretty rapidly towards and this is where the speculation starts and this works well with this kind of framing that vitalic has shared around defensive techno techno optimism as you can see in vitalic view there's both abundant and delightful Futures ahead that are possible as well as very very scary ones and so you can see the prompt here is how do we think about tamper proofing a solved world a world where we do have just much abundance thanks to machine intelligence and so I think this is really the opportunity ahead and for people who are kind of clued into this view now recognizing that the capacity to build applications that interweave systems that create a space a portal a method for individuals to control their own data their own secrets with intelligence associated with their own data and secrets before that agent goes and interacts with the rest of the web really feels like the best possible outcome for the machine intelligence world that we're heading towards because the alternative is basically just a much scarier and Wilder version of what happened with web 2 which is people interacting with wal Gardens they don't have fundamental control over their data and therefore are prone to manipulation which we have continued to see and so we're we do have prizes uh for the hackathon this weekend they are oriented towards decrypting items inside of lit and using them in this agentic context there's also a prize for for Bitcoin is related um but I just want to just reiterate this kind of core idea that it really feels like the opportunity is now to give people access to their own data and making that data really really useful whether it's in the context of Finance like building an automated method for investing in derivatives uh with the uh technical analysis done in an automated way or in the context of user owned data for things like um smarter recom commendations we've just seen some really fascinating work happening from people who have kind of taken this as a fundamental perspective and we'll be be here all weekend and so with that you can find me and you can find us online here thank you so much for your attention and yeah looking forward to building together [Applause] [Music]