Audits Are Not Enough | Odysseus Lamtzidis - phylax systems

[Music] hello everyone uh my name is adicus I'm the CE of phac systems and today we're going to talk about why audits are not enough so the presentation has three points uh one is why 2025 will be a water Set uh um moment for security why security is now moving away from being an afterthought and into uh the center of uh adoption what are the shortcomings of industry and then how we can define a different protocol for hug prevention that actually uh works so Security in 2025 from 2019 approximately to 2025 we had what we call D and capital right we had all these inflows from people that wanted to take a lot of risk uh in 2022 we uh we were about two two uh $200 billion um in crypto assets but the same time we had about $3 billion lost only in 2022 and it's digon right and by digon Capital we mean people that are ready to take risk right they only care about maximizing the yield they they they uh treat crypto something that is inherently risky and they really don't care if you know the UI or the DBS are experimental if things sometimes break if uh security is not top notes as you would expect from uh cyber security best practices uh is what we call the D Capital but that's coming to an end right from this year and on we starting to see more and more uh institutional adoption as uh they're searching for higher and higher yields that uh the traditional Capital traditional Financial system can support uh and they're looking for something that they can't get from treasuries right the problem now is that right now the treasuries uh for at least a 10 year uh United States Treasures is about 4% which means that it's harder and harder uh every year for institutions to uh to explain why they want to take this risk uh in defi when actually you can get pretty safe yield in the traditional system and this type of user this new user Persona is inherently risk averse it's traditional institutions um like black rock or Banks but also it's everyday people that will move from web 2 to web 3 to try new apps to maybe save their uh the livelihoods right and this new cap will be inherently uh more risk adverse they don't uh they don't care about the highest yield per se they care about the higher risk adjusted yield right they want to take opportunities but also manage the risk nevertheless the market is exploding right we're seeing over 110 billions in stable coins across all blockchains uh we had this year about $3 trillion do in transaction volume and in 2024 we had about 44 billions of inflows right so Market is exploding D capital is running out a new kind of capital more institutional more risk aair is joining so the new entrance demand security uh black Gro launches a fund uh which called bidle to start tokenizing assets something that is also echoed by Joseph poock from Bas this new exploration into tokenizing assets on the blockchains we see Kraken and Bas launch uh super chain l2s where not only they want to move a lot of their business from offchain to onchain but if you see in both of these layer to messaging security is a core part right the kren has talked a lot about sequencer based security base is again and again saying that they want to create not only a new uh market for the global economy but also something that is inherently safe and why are these two players talking so much about security is because they understand what I said before that a new entrance other institutions or everyday people are risk averse right they don't uh they can't live with the fact that they may lose their entirety of their Capital just to get 20% yield um and and it's interesting also that other L ones that sui and Aptos are also um using security as a value proposition you know for example for the move language right so we're seeing a uh a move towards security not being something that only is concerned by for Auditors or Builders but it's becoming a core uh marketing position for this new uh networks so what is the security flywheel um as we see more institution adoption we see more inflows right more Capital flowing into the system the crypto system more inflows means that we have not only we have more users on boarding but also uh crypto is moving away from the Twilight and it's coming to the center of attention right there is more trust so trust plus security practices best practices means that Builders can experiment more they can move away from using the same for datab um code bases the same uh mechanism and can start experimenting more because now they have users and if they have also more security they feel they can experiment without feeling that their career will be over also trust and security uh improve the self- regulation of the space right this is something very important um there is a huge discussion right now happening in the space about all the griefs and scums that exist and self-regulation is something very important it's very uh it's critical that we as an industry we can self-regulate and we can show that we can actually apply best practices in security uh because if we don't then we will be regulated violently and unilaterally and finally more security means that we have fewer Black Swan events right uh the bybit hack for example where $1.5 billion were evaporated is a Black Swan and the industry will not be able to take a lot of this so uh more security means that we have fewer and fewer of these incidents so I hope I impressed the importance of security and why it's something that not only us as Builders uh developers but also institutions understand let's talk about the current situation in security so in general we we like to say that security has layers right it's the Swiss ches model or you can think of it like a castle with multiple layers of defenses um we have of course the tests where the developers are responsible for writing there com in many flavors then we have something we call formal verification where we can formally prove uh ma you know in a mathematic way that the code is secure Audits and finally monitoring and in the response right what happens when there is a security incident and how does your team respond to that it's not enough the tests are static you know you write them once and then you have to period periodically update them and they are imperfect right developers forget test cases all the time because it's very hard to find all the Ed cases into how your DB might enter into an incorrect State formal verification is limited because if the protocol is too too big it's very hard to do and also it's hard to do it's something that requires a lot of development effort um which hardly seems a lot of times important um versus Futures then we have audits which are expensive this is the running joke of our industry where protocols pay like uh half uh you know half a million dollars per year for Audits and also they are periodic right you do an audit here an audit there but in the meantime you might deploy small improvements in your code which you think it's safe but a lot of times this uh create vulnerability and finally monitoring an instant response it's too late one of the core value propositions of blockchains is the mutability right when when something happens when a transaction goes on the network it's done so monitoring and inst the response comes from web 2 from uh it's useful to know but in crypto there's nothing you can do about it right you just see your house being uh burned so let's imagine a future where the base layer is actually incentivized to stop hacks before they uh they realized right let's try to imagine how we could have this feature and to do that we have to answer these three questions first how do we Define a hack what is a hack uh which is also different for every DB how can the base layer prevent it right so we Define the hack how can the base layer prevent it and then finally we know what the hack is we found a way for the base layer to prevent it how do we incentivize the base layer the network nodes to actually do it right because it's it's a it's a network you can't force them to do it so this is the core pillars of what we're building at FX and basically we try to answer these three questions how do we Define a hack let's say that you can define a hack um in in bite code right in solidity and you say that I Define an invariant a security Rule and if this thing reverts um then the state transition is malicious the transaction is not right so now that we have Define the haug then we have to find a way for the base layer to commit to enforce it so what do we do it we take this definition and we put it on chain and then the uh the base layer only needs to fetch the definition while it's building the block and say I will not allow any transaction to go into the network which violates this definition and because this definition has come from the DB itself it remains credibly neutral right the the base layer doesn't make any decisions they just respect the wishes of the DB and finally how do we uh incentivize the base layer it's very simple since we have all these things on chain we can create a mechanism and we can say that I will pay you fees to uh as a d i will pay fees to then Network note let's say the sequencer to prevent the hacks um and if you don't do it I'll slash you or you know I'll just use a te so these are the three building blocks of what we're building first we Define the hack as in solidity then we put it on chain and the base layer then the sequencer runs a side car to fetch the definition and continuously check that new transaction doesn't violate the definition and finally the DBS pay the sequencer some extra fees for providing that service this is an example of uh a hack definition or an assertion as we call it this is from the oiler hack and here um it's actually very interesting because this CVM bite code the solidity is executed off chain during block production we can actually do a lot of interesting things for example here what we say is we want to take the state before the state transition and then we want to verify some invariant right we want to make a check and this is a simple invariant right in another more complicated we could also say take the state before the state transition the transaction do something and then take the state after the state transition after the transaction and do something else and if this whole computation reverts that mean the transaction no longer should not be included in the blog and since we're using solidity it's very simple for developers to write it since we're using cheat codes to add a uh new functionality this reminds a lot fundry uh the testing framework so it's easy for developers to be on boarded this is an example of how what assertions are so assertions are this solid the bite code and they have two interesting very uh properties they're Atomic right they they other uh the other are true or false and they're and they read only the state right they don't do anything on the state they they are read only which means that for example wait this transaction uh this transaction call three contracts there are two groups of assertions uh for contract one and contract two and then what we do is after the state transition after the transaction is applied to the state we can also run the relevant assertions and because they are Atomic and read only they can be run in parallel right so what we're doing effectively is we're upgrading the state transition function and we add some functionality some very something very specific that because of its nature we can parallelize very aggressively and this becomes very fast which is why also these uh assertions are better to be uh defined in this specific format than be you know added directly to the smart contract code not only we can do more things but we can do that we can do them much faster so the first integration we're working with uh we're integrated with the op op stack because it's very simple and we're using RAB boost which is a component that allows an OP stack sequencer to have an external block Builder um and what we've built basically is this external block Builder which behaves exactly the same um as the reference implementation but also during blog production it will fetch the relevant assertions and it will check that every transaction doesn't violate them right um and it's actually very important that we're using Mora boost um from flbs in because in this way we're not only it's not easy it's not only easy for networks to integrate but also we are super chain compliant right we we we work on top of the op stack um and this is something very important because we don't want to violate uh the underlying protocol we want to add some new functionality so what's the conclusion for this presentation very quickly One Security will play a central role uh onward will stop being something that is only relevant for security Engineers or Builders but will be relevant for everyone as we on board this new type of users it needs to be preventative and also it needs to be dynamic it doesn't audits are not enough this periodic check of the Environ is not enough um we can build new Solutions uh that solve soled almost upgrade the underlying protocol um you know like Lia for example uh or FX why Li do exist it exists because for whatever reason the ethereum protocol doesn't offer native raking uh Native delegated stake right in the same way we the file rable layer gives this new functionality to the underlying protocol it enables a protocol to check for invariants which are defined uh by the by the abbs themselves by Builders right why this is not part of the protocol for reasons because the protocol can't be can be can can't evolve and shouldn't evolve fast we can have also very powerful preventative Solutions um by building a mechanism without needing invasive s Solutions or you know AI black boxes um and finally we can make security intuitive by reusing the mental models that people know and this last part is very important because security needs to be intuitive not only for the people who consume it like the end users who want to understand what the security properties of that DB they will use but also for the builders that will have to use the security tool right if we if we force them to use some very complicated new language or new Frameworks they wouldn't adopt it but but by reusing the mental models they have and this is just generally very good product philosophy you can build new Concepts you can build new tools new functionality with a minimum amount of uh onboarding difficulty and this is for example the the MVP of the UI where anyone can observe the project a contract and then the relevant assertions right um which is very important to keep this protocol neutral users can inspect what envirment have been added to a protocol not only to know the expected Behavior but also so they can know that they don't have to trust the dub for doing the best practices in terms of security but instead uh they can verify themselves so let's solve uh security so crypto can win because that's theend the important part at the end of the day it's not about um building interesting security tooling it's about making crypto win um thank you [Music]