Verifiable Computation and Multi-Party Workflows | Fernando Castillo - Dober

[Music] hello everyone thank you for being here I'm Fernando Castillo and I'm going to be talking about verifiable computation and multiparty workflows first uh a bit about me I'm a startup founder I'm currently a researcher at eberlin where I'm also doing my PhD and I'm also a Cofe drink care so first a bit about the motivation for this is that we want to create a workflow that uses the input from multiple parties and that all the parties that are involved can trust in each other works that would mean then to have trust verification and auditability what could be a specific use case for this uh decentralized physical uh infrastructure Network where you have the individ idual use for the asset and at the same time the asset accounting uh let's say we have an EV charger that produces um specific and a specific amount of energy every day and at the same time it's used to load specific amount for each electrical vehicle so in an ideal scenario we would be have having only the periodical information aggregated and every user would have the confidentiality of the use of their asset so for this we would need the different properties one of them would be confidentiality meaning that only the person the The Entity that needs to know about the information would know and verifiability for instance in the example of the the accounting that all the some on the agregation of the use it it's match up and interoperability when you have the multiple parties uh talking to each other so this would lead to multiple questions one of them would be how can the parties then keep the confidentiality in each of the steps another question would be how can the parties do the verification in the whole workflow so for this we will have different tools um we will call verifiable computational components one of them Ser knowledge proofs and another secure multiparty computation and also trusted execution environments uh but all of them had its pros and cons for instance C knowledge Pro could ensure the computation Integrity but we will require that the execution of the Ser knowledge program know about the inputs but at the same time they will guarantee the Integrity of the computation and the output of the cryptographical proof they generate also we we have in Secure multiparty computation that the confidentiality could be kept at the same time for the input that's been uh delivered for all the parties that are doing the computation and but here we will trust that the computation is handled correctly for all the parties that are involved and in the case of the tees we could have the confidentiality for the result of the of the computation that's executed and for the input that's given to the T because in Te You would have uh not access from the even for the owner of the machine that's running the code but here we wouldn't have computational Integrity guaranteed in terms of a cryptographical proof like with C knowledge proof but here we would have attestation mechanisms that are built in in the hard World itself so we have all of these Technologies what would happen if we would mix them together then we would create this type of boxes where you would have a mixes properties of confidentiality and at the same time the verification of the proofs that they could generate for instance if you mix C knowledge proofs with a a te you could guarantee that the data that's being delivered to the te is only confidential and not even the machine that's running the code could access the information that's being given and at the same time you could generate a cryptographical proof of of the computation that's being made and hence you could generate workflows like the one Illustrated here so this would require a specific setup but in and in which we would require to register all the keys for the each verifiable components for instance uh if we would have a Zer knowledge uh like s snarks we would need to register the verification keys that are being use to verify the computation and at the same time we would need to include the previous keys and the current box that we're using for our VCC so in the example that was discussed before we would have the electrical vehicle Chargers where they could do an authentic an anonymization in the authentication process where a user uses the the electrical vehicle charger and at the same time then we would have at an aggregation verifiable computational component and finally this could be verified in the blockchain for public transparency and for the tokenomics that this could deliver so this scheme could also benefit from an an an update in the different components that are there and in this case we would only need to update the registry in which the keys are registered and only do an update in the vccs that has have direct dependency of the previous in the workflow so what's the vision for this type of scheme is where the a workflow in which you have could have multiple parties working together and each of them having the confidentiality and the trust in each other works and all of this could be implemented for instance in in like in this example that was described before you could have the electrical V ver electrical vehicle charger on one side and the electrical vehicle on the other and at the same time the owners of those assets could guarant be guaranteed that the use of the their assets has been correctly computed for accounting terms but this might might sound that it's a kind of kind of salt but currently there are a few challenges especially these four challenges here one is the high or higher computational cost and overhead for composing or nesting also there's a difficulty when using uh elliptical cryptographical curves because they need to March if you have a composition of proof one inside of the other and as described before this also requ require a specific setup that might be difficult or complex to implement with multiple parties so the current Focus for this is uh the decentralization of electrical vehicle charges thus we could enable on a firstep real world asset tokenization and at the same time increase the confidentiality in the workflow that users are being put into so as other type of application for this type of of Technology would be authenticated machine learning Federated or CIP learning where you would require a uh training in a specific data set that's authenticated or also we could use this for supply chain certification for instance to calculate or compute a full CO2 uh footprint emission for a specific product or to generate green hydrogen certificates well thank you very much I I will be around here thank you