Decentralized ai agent network system built upon blockchain-based web3 ai-id and web3 ai-contract for verifiable confidential ai agent computing

Abstract

The apparatus comprises a first web3ai-ID, a memory storing a private key corresponding to the first web3ai-ID, a communication interface for communicating with a blockchain network in which a first web3ai contract corresponding to the first web3ai agent is registered, and at least one processor, wherein the at least one processor generates a pair of processor private keys and processor public keys, transmits the processor public key outside the at least one processor, executes code corresponding to at least one AI agent service specified in the first web3ai contract, obtains a code hash value corresponding to the executed code, obtains a state root value for personal data processed through the executed code, generates a first electronic signature for the code hash value and the state root value based on the processor private key, and transmits a registration request for the first electronic signature, which includes the first electronic signature and a second electronic signature generated based on the private key corresponding to the first web3ai-ID, to the blockchain network so that the first electronic signature is registered in the first web3ai contract. A first web3ai agent is disclosed. In addition to this, various embodiments identified through the specification are possible.

Description

Technology Field

[0001] The embodiments disclosed in this document relate to AI agent technology, blockchain technology, and WEB 3.0 technology. Background Technology

[0002] Prior to the AI ​​era, software was implemented to operate on Web 2.0 systems centered on client applications and central servers operated by service providers. Meanwhile, with the continuous advancement and performance improvement of AI technology, there is a growing need for AI agent models capable of independently assessing situations and taking appropriate actions without direct user intervention or instructions. Once AI agents become widespread, individual users and existing web service providers will each possess autonomously operating AI agents. Such an environment is difficult to create within the existing central server-centric, cloud computing-based SaaS paradigm. The problem to be solved

[0003] If AI agents are implemented by a specific company, personal data or the AI ​​agents become the property of that company, potentially infringing upon the sovereignty over personal data. Therefore, we aim to provide a blockchain-based network protocol that enables the implementation of decentralized, autonomously operating AI agents that are not owned by any specific platform. By utilizing blockchain-based Decentralized Identifier (DID) web standard technology for AI agent IDs, we intend to allow users to retain sovereignty over their IDs and data regarding the AI ​​agents by using blockchain-based DIDs generated on user devices, rather than relying on the central server systems of specific companies. Furthermore, by enabling identity verification based on blockchain even when AI agents operate autonomously in a decentralized manner, we aim to ensure that AI agents can communicate with one another in a trusted environment.

[0004] A large amount of personal data is required to enhance the performance of AI agents. Consequently, serious side effects, such as the leakage or infringement of private information, may occur during the processing of large volumes of personal data. For example, server devices must be used to process massive amounts of data, and there may be a risk of data leakage during the communication process with these server devices. Therefore, by establishing a mechanism to prevent the leakage of personal data within the aforementioned protocol, we aim to provide an AI agent that can manage and process massive amounts of personal data without the risk of data leakage.

[0005] In addition, in order to enable AI agents to automatically perform various tasks, it is necessary to unify the interfaces among service providers of services for AI agents (hereinafter referred to as AI agent services). If different interfaces are used by each service provider, the scalability of the AI ​​agent services and, furthermore, the scalability of the AI ​​agent ecosystem is limited. Accordingly, this invention aims to provide a protocol capable of scalability by disclosing the AI ​​agent service interface based on blockchain and encouraging the reuse of the disclosed AI agent service interface among different AI agents. Hereinafter, the AI ​​agent implemented on the protocol disclosed through this invention is referred to as the web3ai agent. means of solving the problem

[0006] A first web3ai agent according to an embodiment disclosed in this document comprises a first web3ai-ID, a memory storing a private key corresponding to the first web3ai-ID, a communication interface that communicates with a blockchain network in which a first web3ai contract corresponding to the first web3ai agent is registered, and at least one processor, wherein the at least one processor generates a pair of processor private keys and processor public keys, transmits the processor public key to the outside of the at least one processor, executes code corresponding to at least one AI agent service specified in the first web3ai contract, obtains a code hash value corresponding to the executed code, obtains a state root value for personal data processed through the executed code, generates a first electronic signature for the code hash value and the state root value based on the processor private key, and receives a request for registration of the first electronic signature including the first electronic signature and a second electronic signature generated based on the private key corresponding to the first web3ai-ID, wherein the first electronic signature is the first It is configured to be transmitted to the aforementioned blockchain network so as to be registered in the web3ai contract. Effects of the invention

[0007] In a protocol according to an embodiment disclosed in this document, web3ai agents can verify identity information through a blockchain network and autonomously communicate based on mutual trust. In addition, through the interface of the AI ​​agent service disclosed through the blockchain network, the AI ​​agent services provided on the protocol can be provided based on a unified interface and have scalability.

[0008] A web3ai agent according to one embodiment disclosed in this document can provide personalized services based on vast amounts of personal data while guaranteeing to users that there is no risk of personal data leakage. In addition, various effects that can be identified directly or indirectly through this document may be provided. Brief explanation of the drawing

[0009] FIG. 1 is a diagram for explaining a protocol provided through a blockchain network and a plurality of web3ai agents according to one embodiment. FIG. 2 is a block diagram of a web3ai agent according to one embodiment. FIG. 3 is a diagram illustrating a method to prevent the exposure of personal data processed on a web3ai agent according to one embodiment. FIG. 4 is a signal flowchart of a method for updating a TEE / ZK rollup contract and verifying a TEE digital signature according to one embodiment. FIG. 5 is a signal flow diagram of a method in which an AI agent service interface is created and used on a protocol according to one embodiment. FIG. 6 is an example of a product recommendation service provided by a plurality of e-commerce service web3ai agents according to one embodiment. Figure 7 is an example of an interface for a product recommendation service associated with an e-commerce service. In relation to the description of the drawings, the same or similar reference numerals may be used for identical or similar components. Specific details for implementing the invention

[0010] Hereinafter, various embodiments of the present invention are described with reference to the accompanying drawings. However, this is not intended to limit the present invention to specific embodiments and should be understood to include various modifications, equivalents, and / or alternatives of the embodiments of the present invention.

[0011] FIG. 1 is a diagram for explaining a protocol provided through a plurality of web3ai agents (100, 200, 400) and a blockchain network (300) according to one embodiment. Referring to FIG. 1, a personal web3ai agent (100, 200) and a service web3ai agent (400) are illustrated as examples for explaining the plurality of web3ai agents (100, 200, 400). The plurality of web3ai agents (100, 200, 400) can be classified into a personal web3ai agent (100, 200) and a service web3ai agent (400) depending on whether the web3ai agent owner (operator) is an individual or a service company. The personal web3ai agent (100, 200) and the service web3ai agent (400) may have the same technical configuration and perform the same functions, differing only in their roles in the protocol.

[0012] In a protocol according to one embodiment of the present invention, web3ai agents can communicate with other web3ai agents disclosed on the internet and operate in cooperation. A plurality of web3ai agents (100, 200, 400) of FIG. 1 can communicate with each other and be connected through a network to operate together.

[0013] In the protocol, each of the multiple web3ai agents (100, 200, 400) can be identified by a web3ai-ID. Each web3ai agent (100, 200, 400) can have one web3ai ID. The web3ai ID can be understood as a unique identifier of the web3ai agent in the protocol.

[0014] According to one embodiment, the web3ai-ID can be implemented as a blockchain-based Decentralized Identifier (DID). Multiple web3ai agents (100, 200, 400) can directly generate the web3ai-ID. Multiple web3ai agents (100, 200, 400) can generate a pair of public key / private key corresponding to each web3ai-ID. The said pair of public key / private key can function as an account on the blockchain network (300). According to one embodiment, the web3ai-ID may be the public key and may be a separate DID corresponding to the public key.

[0015] Multiple web3ai agents (100, 200, 400) can generate a digital signature using the private key and can authenticate ownership of the web3ai-ID through the digital signature. Additionally, multiple web3ai agents (100, 200, 400) can create a confidential communication channel with other web3ai agents through data encryption using the private key.

[0016] Hereinafter, a web3ai agent (100) is described as an example of a plurality of web3ai agents (100, 200, 400). The description of the web3ai agent (100) can be applied equally to other web3ai agents (200, 400).

[0017] A web3ai agent (100) can create a web3ai contract (310) corresponding to its web3ai-ID on a blockchain network (300). The web3ai agent (100) can create a web3ai-ID and the private key and public key corresponding to the web3ai-ID, and send a request to create a web3ai contract corresponding to the web3ai-ID to the blockchain network (300). The request to create a web3ai contract may include an electronic signature generated based on the web3ai-ID and the private key corresponding to the web3ai-ID. A web3ai contract corresponding to the web3ai is created on the blockchain network (300). In order for anyone to verify the web3ai agent (100) through the web3ai contract (310), the web3ai agent (100) can be made public by registering information related to itself in the web3ai contract (310). When a web3ai agent (100) transmits a request for information registration to a blockchain network (300), which includes information to be registered and a digital signature generated based on a private key corresponding to a web3ai-ID, the information can be registered in a web3ai contract (310) corresponding to a web3ai-ID.

[0018] Existing smart contract technology is limited to registering a smart contract interface and operation code on a blockchain network and enabling their execution. In the web3ai contract (310) according to one embodiment of the present invention, by extending existing smart contract technology, a DID document (311), a public certificate (312) issued for a web3ai-ID, an off-chain AI agent service endpoint (313), an off-chain AI agent service interface (314), an on-chain AI agent TEE / ZK rollup contract (315), and an on-chain smart contract (316) can be registered.

[0019] In various embodiments, the web3ai contract (310) enables verification of information related to the web3ai agent (100) operating off-chain or actions performed by the web3ai agent (100) through the blockchain network (300). Hereinafter, "off-chain" may be understood as a term referring to an external network that is not on the blockchain network (300).

[0020] In various embodiments, data included in the items (311, 312, 313, 314, 315, 316) of the web3ai contract (310) can be registered or modified only by a request including a digital signature generated based on the private key corresponding to the web3ai-ID of the web3ai agent (100). Accordingly, only the web3ai agent (100), who is the owner of the web3ai-ID, has administrative authority over the web3ai contract (310).

[0021] A DID document (311) according to one embodiment may contain information associated with a web3ai-ID. For example, a public key corresponding to a web3ai-ID may be stored in the DID document (311). The public key may be understood as a public key that can be controlled by the owner of the web3ai-ID. In various embodiments, a processor public key generated by the processor (110) of the web3ai agent (100) may be stored in the DID document (311).

[0022] A public certificate (312) according to one embodiment can be understood as a list of public certificates in the form of Verifiable Credentials (VCs), which are web standard (W3C standard) specifications issued for a web3ai-ID. A public certificate (312) of a web3ai contract (310) can be understood as a list of VCs disclosed through a blockchain network (300) among the VCs owned by a web3ai agent (100). A web3ai agent (100) can register the VCs it intends to disclose among the VCs it owns in the public certificate (312) of the web3ai contract (310). The owner of a web3ai-ID can optionally disclose information associated with the owner of the web3ai-ID through the public certificate (312) of the web3ai contract (310). For example, a public certificate may contain personal information (e.g., identity information) about the owner of a web3ai-ID issued by trusted authorities (e.g., government agencies, companies, medical institutions, banks, etc.). In various embodiments, the trustworthiness of a web3ai-ID that possesses a large amount of personal data in the form of a VC issued by trusted authorities may be enhanced.

[0023] In various embodiments, at least some of the personal data stored by the web3ai agent (100) (e.g., personal data (134) of FIG. 2) may be managed in the form of a VC. The VC includes the digital signature of its issuer and the web3ai-ID of its owner. Thus, the source of the personal data (the VC issuer and data creator) and the web3ai-ID (the holder of the VC and data owner) can be bound. Personal data managed and processed in the form of a VC is tamper-proof, and its source and owner can be verified through a blockchain network (300). In the above protocol, the identity of the web3ai agents can be authenticated through the blockchain network (300) (e.g., through an identity authentication VC), so users of the web3ai agents can be safely protected from deep-fakes, spam, voice phishing, etc.

[0024] In various embodiments, the web3ai-ID owner may register only some of the VCs issued for the web3ai-ID in the public certificate (312) of the web3ai contract (310) and store the rest only in the web3ai agent (100). In some embodiments, the VCs not disclosed in the public certificate (312) may be provided directly to the web3ai agent in the form of a Verifiable Presentation (VP) containing at least some information contained in the VC and a digital signature generated based on the private key corresponding to the web3ai-ID when requested by another web3ai agent.

[0025] The off-chain AI agent service endpoint (AI agent Service Endpoint (ASE), hereinafter ASE) (313) can be understood as information about where the web3ai agent (100) is being serviced. For example, the ASE (313) may include the URL of a mobile device (e.g., edge agent (10) in FIG. 2) or a cloud server (e.g., cloud agent (20) in FIG. 2) that is implementing the web3ai agent service. Through the service endpoint registered in the ASE (313), other web3ai agents can access the web3ai agent (100).

[0026] The off-chain AI agent service interface (AI agent Service Interface (ASI), hereinafter ASI) (314) is information about what service functions the web3ai agent (100) provides, and may include specifications of AI agent services being provided off-chain. The web3ai agent (100) may register specifications of services to be provided to other web3ai agents on this protocol in the ASI (314). Specifications of AI agent services may include interface types such as those exemplified in Table 1.

[0027] AI agent Service Interface Types Natural Language Chat Interface, Audio / Video Call Interface, Remote Procedure Call (RPC) Interface, Publish-Subscribe-Notify Interface

[0028] In various embodiments, the ASI (314) may include specifications disclosed by other web3ai agents (e.g., web3ai agents (200, 400)). For example, if a second web3ai agent (200) discloses a specific AI agent service interface through a second web3ai contract (320), the first web3ai agent (100) may import the disclosed AI agent service interface of the second web3ai agent (200) into the ASI (314) of its first web3ai contract (310). In this way, the AI ​​agent service interface disclosed through the web3ai contract of the blockchain network (300) can be imported and reused by other web3ai agents.

[0029] In various embodiments, multiple web3ai agents (100, 200, 400) can autonomously use AI agent services disclosed by different web3ai agents through ASE and ASI disclosed on web3ai contracts (310, 320, 340) registered on the blockchain network (300), provide the same service through a common interface, or implement various AI agent services for users by linking AI agent services with each other.

[0030] Through the DID document (311) and public certificate (312) registered in the web3ai contract (310), the DID information and public identity information of the web3ai agent (100) can be disclosed on the blockchain network (300). Accordingly, anyone can search for the disclosed web3ai agent (100) based on the web3ai-ID and look up the identity information of the web3ai agent (100). In addition, anyone with access rights can call the service provided by the web3ai agent (100) through the ASI (314) or communicate with the web3ai agent (100).

[0031] According to one embodiment, an implementation (hereinafter referred to as an AI agent service implementation) that enables various AI agent services provided through ASI (314) may be included in an edge agent (10) and / or a cloud agent (20) included in the web3ai agent (100). An on-chain AI agent TEE / ZK rollup contract (hereinafter referred to as a TEE / ZK rollup contract) (315) may include evidence data that can verify the operation of AI agent service implementations executed off-chain. For example, evidence data regarding whether the AI ​​agent service is performed as disclosed to ASI (314) may be periodically registered in the TEE / ZK rollup contract (315).

[0032] In various embodiments, since the web3ai agent (100) is inevitably driven based on the user's private data, it is necessary to ensure a confidential computing environment in which private data is not exposed during the operation process. A TEE / ZK rollup contract (315) can perform the role of ensuring this. For example, a Remote Attestation (RA) digital signature allows verification that an AI agent service operating off-chain has been executed on the processor's TEE mode by a confidential-containerized program. Additionally, a ZK attestation allows verification that an AI agent service has been executed on the zkVM. A detailed explanation of this will be provided later through FIGS. 3 and 4.

[0033] On-chain smart contracts (316) can be understood as functions of existing smart contracts. On-chain smart contracts (316) can be understood as smart contract code deployed on a blockchain network (300). On-chain smart contracts (316) can support traditional smart contract technology that enables functions such as issuing tokens or trading tokens (DEX) to be executed on the blockchain network (300). For example, an AI agent service can be set up as a paid model, and the usage fee for the AI ​​agent service can be settled through tokens on the blockchain network (300).

[0034] The protocol on which the web3ai agents disclosed in the present invention operate is difficult to run on existing Web 2.0 ID systems. Existing ID systems may include, for example, ID systems for IDs / passwords stored / managed by central servers operated by service providers, such as for-profit companies (e.g., Google, Apple, Meta, OpenAI), or ID systems such as OAuth-based SSO. If web3ai agents are implemented on Web 2.0 ID systems, ownership and control of individuals / companies' web3ai agents are transferred to a specific company's platform. Consequently, the private data required to operate the web3ai agents is also owned by that specific company, and there is also a risk that private data may be leaked by that company's central server.

[0035] Through the web3ai-ID and web3ai contract (310) according to one embodiment, a blockchain-based decentralized and verifiable computing environment can be created. In this environment, an open AI agent ecosystem can be established in which multiple web3ai agents (100, 200, 400) can operate in conjunction with each other. For example, when multiple web3ai agents (100, 200, 400) want to communicate with each other or use a service disclosed by the other party, they can first look up the identity information (e.g., Identity Verification VC) of the web3ai agent registered in the public certificate of the web3ai contract (310, 320, 340) on the blockchain network (300). Alternatively, when multiple web3ai agents (100, 200, 400) connect to a specific web3ai agent, they can directly request identity information (e.g., Identity Verification VC) from the connected web3ai agent in real time and verify the identity. Through this, multiple web3ai agents (100, 200, 400) can perform data communication based on a mutual trust relationship.

[0036] In various embodiments, web3ai agents can generate the same cryptographic key through key exchange with each other (e.g., Diffie-Hellman key exchange) and perform secret communication using said cryptographic key.

[0037] FIG. 2 is a block diagram of a web3ai agent (100) according to one embodiment. The web3ai agent (100) may include at least one processor (110), a memory (120), and a communication interface (140).

[0038] Referring to FIG. 1, the web3ai agent (100) may include an edge agent (10) and a cloud agent (20). The edge agent (10) can be understood as an agent operating on a mobile device, such as a smartphone, of the web3ai-ID owner (user). The cloud agent (20) can be understood as an agent operating on a cloud server that can only be accessed by the web3ai-ID of the web3ai agent (100). The edge agent (10) and the cloud agent (20) interact with each other through a network to operate as a single web3ai agent (100). The edge agent (10) and the cloud agent (20) may each include the components of the web3ai agent (100) shown in FIG. 2.

[0039] The edge agent (10) according to the various embodiments disclosed in this document may be a device of various forms. The edge agent (10) may include, for example, a portable communication device (e.g., a smartphone), a computer device, a portable multimedia device, a portable medical device, a camera, a wearable device, or a home appliance. The edge agent (10) according to the embodiments of this document is not limited to the aforementioned devices.

[0040] The processor (110) of the web3ai agent (100) may include at least one processor that controls the operation of the web3ai agent (100). The processor (110) of the web3ai agent (100) may be connected to memory (120) so as to be operable and configured to execute instructions stored in memory (120). In various embodiments, the processor (110) of the cloud agent (20) may include a high-performance processor such as a GPU or an NPU.

[0041] In one embodiment, the processor (110) of the web3ai agent (100) can generate a pair of private and public keys (processor private key, processor public key) based on a known encryption algorithm (e.g., RSA encryption algorithm) on a hardware security module (e.g., Trusted Execution Environment (TEE)) mounted on the processor (110). At this time, the processor private key may be stored in the security module within the processor (110), and the processor (110) may transmit the processor public key outside the processor (110). The processor (110) may include a security module, and the processor private key may be stored in said security module.

[0042] In various embodiments, the processor public key transmitted outside the processor (110) may be stored in the web3ai-ID / key information (132) of the web3ai agent (100), the DID document (311) of the web3ai contract (310), and / or the TEE / ZK rollup contract (315).

[0043] In various embodiments, the processor (110) may generate a new processor key at a set period. If the processor key is changed, the changed processor key may be additionally stored in the web3ai-ID / key information (132), the DID document (311) of the web3ai contract (310), and / or the TEE / ZK rollup contract (315). Based on the processor public key disclosed through the blockchain network (300), the TEE digital signature may be verified.

[0044] The memory (120) of the web3ai agent (100) may include an operating system (122), web3ai-ID / key information (132), personal data (134), a confidential-containerized program (136), and an AI model (138).

[0045] The operating system (122) may include an AI agent module (124). The AI ​​agent module (124) can be understood as an artificial intelligence model implemented to autonomously control the web3ai agent (100) on behalf of the user of the web3ai agent (100). The AI ​​agent module (124) may be embedded in the operating system (122) and can control the operation of the web3ai agent (100) (the operation of the processor (110)) at the level of the operating system.

[0046] The web3ai-ID / key information (132) may include a web3ai-ID generated by the web3ai agent (100) and a pair of private / public keys corresponding to the web3ai-ID. In another example, if there is a processor private / public key generated by the processor (110), the web3ai-ID / key information (132) may store the processor public key.

[0047] The personal data (134) of the web3ai agent (100) can be understood as private information associated with the owner (user) of the web3ai-ID. The personal data (134) may be collected from the edge agent (10) or downloaded from a third-party service. The personal data (134) may be generated based on data generated while the user uses the edge agent (10) or data generated while the user uses a third-party service (e.g., social media service, internet search service, etc.). The web3ai agent (100) may provide a personalized response to the user based on the personal data (134).

[0048] In various embodiments, personal data (134) may include multiple data sets, and the multiple data sets may have a unified format. For example, personal data (134) may be stored in an integrated format that is verifiable.

[0049] In various embodiments, personal data (134) may be stored in an edge agent (10) and a cloud agent (20). The personal data (134) stored in the cloud agent (20) may have a large capacity as massive data. The personal data (134) stored in the cloud agent (20) may be encrypted for security. For example, the personal data (134) may be encrypted based on a private key corresponding to a web3ai-ID. The encrypted personal data (134) may be broken down into multiple pieces and then distributed and stored across multiple data centers. Through this, the personal data (134) can be managed so as not to be dependent on a single cloud infrastructure service. When the personal data (134) is encrypted using a blockchain DID (web3ai-ID), only the owner of the web3ai-ID can access the data. Furthermore, because it is encrypted, personal information leakage cannot occur while the personal data (134) is stored.

[0050] A confidential containerized program (136) can be understood as a program that operates in a confidential state to maintain data security. A confidential containerized program (136) can be understood as a program implemented to provide AI agent services. In various embodiments, the web3ai agent (100) may store an AI agent service program that operates based on personal data (134) in the form of a confidential containerized program (136).

[0051] The AI ​​model (138) may include at least one artificial intelligence model necessary for the operation of the personalized AI agent provided by the web3ai agent (100). For example, the AI ​​model (138) of the edge agent (10) may be understood as an on-device AI model (e.g., SML). The AI ​​model (138) of the cloud agent (20) may be understood as an AI model driven by massive data (e.g., LLM).

[0052] The web3ai agent (100) can communicate with other web3ai agents and the blockchain network (300) through the communication interface (140). The edge agent (10) and the cloud agent (20) can be connected through the communication interface (140) and operate as a single web3ai agent (100).

[0053] FIG. 3 is a diagram illustrating a method for preventing the exposure of personal data processed on a web3ai agent (100) according to one embodiment. FIG. 4 is a signal flowchart of a method for updating a TEE / ZK rollup contract (315) and verifying a TEE digital signature according to one embodiment.

[0054] According to various embodiments, a large amount of personal data is required for the web3ai agent (100) to operate. Since the edge agent (10) cannot store and manage all of the large amount of data, almost all personal data (134) is backed up to the cloud agent (20), and only frequently used data can be stored in the edge agent (10). Additionally, the cloud agent (20) is required to execute AI models such as LLM that require high-performance computing resources (e.g., GPU, NPU, HBM) which are difficult to process with only on-device computing resources.

[0055] In particular, since the cloud agent (20) operates on a server, when the cloud agent (20) processes personal data (134), the personal data (134) may be exposed to the server's memory. In this case, there is a possibility that the personal data (134) may be leaked to cloud infrastructure providers (e.g., AWS, GCP, Azure). Referring to FIGS. 3 and 4, a method to prevent the exposure of personal data when personal data is processed by the web3ai agent (100) (specifically, the cloud agent (20)) will be explained below.

[0056] In one embodiment, the processor (110) of the web3ai agent (100) can generate a processor key (4010) on a hardware security module mounted on the processor (110). The processor key may include a pair of processor private keys and processor public keys. The processor private key is stored within the processor (110), and the processor public key can be transmitted outside the processor (110).

[0057] The web3ai agent (100) can enable TEE mode (4020). For example, when the web3ai agent (100) operates a program based on personal data (134), it can enable TEE mode provided through the processor (110) of the web3ai agent (100).

[0058] For example, Confidential Computing can be performed based on a Trusted Execution Environment (TEE), which is a hardware security module installed in the processor (110) (e.g., CPU / GPU processor) of the cloud agent (20). The TEE mode can be understood as the state in which the security module is in operation. Code and processed data executed in the TEE mode of the processor (110) of the cloud agent (20) can be encrypted by a process private key generated by the processor (110). Since the code and data are processed while encrypted in memory (120), even an operating system (e.g., operating system (122)) capable of accessing the memory of the hardware on the server where the cloud agent (20) is running cannot access the code and data in TEE mode.

[0059] In various embodiments, the code of a program (e.g., a confidential containerized program (136)) driven by the web3ai agent (100) may be made public through an open platform, and the interface of the code (program) may be specified in the ASI (314). The disclosed code may correspond to a single hash value. Hereinafter, the single hash value corresponding to the disclosed code is referred to as the code hash.

[0060] The web3ai agent (100) can execute code corresponding to at least one AI agent service specified in the web3ai contract and obtain a code hash value corresponding to the executed code (4030). The web3ai agent (100) can obtain a state root value for personal data processed through the executed code (4040). The web3ai agent (100) can generate an electronic signature (hereinafter referred to as a TEE electronic signature) for the code hash value and the state root value based on the processor private key (4050). The TEE electronic signature can be understood as an electronic signature capable of remote attestation (RA). When the cloud agent (20) is executed in TEE mode, the hash value of the code currently being executed (code hash) and the electronic signature for the state root value of the personal data (e.g., personal data (134)) used in the code can be generated by the processor (110) of the cloud agent (20), that is, the hardware itself.

[0061] The web3ai agent (100) can send a request for TEE digital signature registration to the blockchain network (300) (4060). The request for TEE digital signature registration may include a web3ai-ID, a processor public key, a TEE digital signature, and a digital signature generated by the private key corresponding to the web3ai-ID. The TEE digital signature and the processor public key may be registered in the TEE / ZK rollup contract (315) of the web3ai contract (310) of the web3ai agent (100) (4070). While the TEE mode is active, the web3ai agent (100) can generate a TEE digital signature at a set interval through operations 4030 to 4070 and send a request for TEE digital signature registration for the generated TEE digital signature to the blockchain network (300). Accordingly, a new TEE digital signature may be registered in the TEE / ZK rollup contract (315) at a set interval.

[0062] Since the TEE digital signature is a digital signature for the code hash of the personal information processing program executed in TEE mode and the state root value of the personal data, it is possible to determine which program was executed through the code hash value. In addition, as the code is executed, the state root value of the personal data changes, so multiple TEE digital signatures registered in the TEE / ZK rollup contract (315) prove that when a specific public code is executed, the code and personal data are not exposed in TEE mode.

[0063] In various embodiments, even if the web3ai agent (100) executes specific code through an external cloud server device, it can be guaranteed that the code disclosed is executed safely without personal information leakage through the TEE digital signatures registered in the TEE / ZK rollup contract (315). Additionally, since the TEE digital signatures are disclosed on the blockchain network (300), anyone can verify and confirm the TEE digital signatures.

[0064] For example, when the second web3ai agent (200) wants to verify the TEE digital signature of the first web3ai agent (100), the second web3ai agent (200) can look up the TEE digital signature and processor public key registered in the TEE / ZK rollup contract (315) of the first web3ai agent (100) (4080). The second web3ai agent (200) can verify which manufacturer's processor the processor public key was generated by using a specific method provided by the manufacturer of the processor (110) (e.g., NVIDIA, Intel). In addition, the code hash value that formed the basis of the TEE digital signature can be obtained through the processor public key. Meanwhile, since the code corresponding to the code hash value is publicly available through an open platform, the second web3ai agent (200) can identify which code was executed by the first web3ai agent (100).

[0065] In various embodiments, the TEE digital signature can be verified by contract code. For example, a TEE / ZK rollup contract (315) may be created so that a request to register a TEE digital signature of the web3ai agent (100) can be verified on the blockchain network (300). When the verification process of the TEE digital signature (e.g., operation 4090) is performed on the TEE / ZK rollup contract (315), the blockchain network (300) may register only the verified TEE digital signatures to the TEE / ZK rollup contract (315), and may reject the request to register the TEE digital signature if the verification fails. If the TEE / ZK rollup contract (315) is implemented to perform TEE digital signature verification, it can be proven that the web3ai agent (100) operates without the exposure of personal data if it is confirmed that the TEE digital signature is registered in the TEE / ZK rollup contract (315).

[0066] In various embodiments, an AI agent service that is open source and free from personal information leakage can be virtualized through Confidential Container technology and executed in the TEE mode of the CPU of the cloud agent (20). The disclosed AI model can be executed in a manner that eliminates the possibility of leakage of personal data used for AI model training and inference through the TEE mode provided by the GPU of the cloud agent (20).

[0067] In another example, the AI ​​agent service of the web3ai agent (100) may be implemented on a Layer 2 blockchain network that operates based on a Layer 1 blockchain network (e.g., blockchain network (300)). For example, the AI ​​agent service may be implemented on a zkEVM (zero-knowledge Ethereum Virtual Machine) or a zkVM (zero-knowledge Virtual Machine) based on the RISC-V instruction set. In this case, the web3ai agent (100) can generate a Zero-Knowledge Proof that the publicly authenticated code was executed successfully without exposing the processed private data, and can periodically register the generated Zero-Knowledge Proof to the TEE / ZK rollup contract (315) of the web3ai contract (310) corresponding to the Layer 1 blockchain network (300). Additionally, since this Zero-Knowledge Proof is public on the blockchain network (300), anyone can verify it.

[0068] It can be ensured that AI agent services are executed without leakage of private data by using the TEE / ZK rollup contract (315) of the web3ai contract (310) and a virtual machine (VM) based on TEE technology and ZK technology of the processor hardware.

[0069] FIG. 5 is a signal flow diagram of a method in which an AI agent service interface is created and used on a protocol according to one embodiment. For example, it is assumed that the first web3ai agent (100) is a private agent owned by an individual, and the second web3ai agent (400) is a service agent owned by a company.

[0070] The second web3ai agent (400) can create an interface for AI agent services (5010) and send a registration request for the interface to the blockchain network (300) (5020). The registration request for the interface may include the interface to be registered, the second web3ai-ID of the second web3ai agent (400), and an electronic signature generated based on the private key corresponding to the second web3ai-ID. The blockchain network (300) verifies the second web3ai-ID that sent the registration request based on the electronic signature and can register the interface to the ASI of the second web3ai contract (340) of the second web3ai agent (400) (5030).

[0071] The first web3ai agent (100) can query the interface of the AI ​​agent service provided by the second web3ai agent (400) through the blockchain network (300) (ASI of the second web3ai contract (340)) (5030). The first web3ai agent (100) can query the ASI of the second web3ai contract (340) based on the second web3ai-ID of the second web3ai agent (400) and select the interface to use (5040).

[0072] The first web3ai agent (100) can call the second web3ai agent (400) according to the selected interface (5060). The first web3ai agent (100) can receive a response to the call from the second web3ai agent (400) (5070). The first web3ai agent (100) can create at least one parameter included in the selected interface. The first web3ai agent (100) can call the second web3ai agent (400) based on the created parameter. The first web3ai agent (100) can query the ASE included in the second web3ai contract (340) of the second web3ai agent (400) and call the second web3ai agent (400) based on the obtained service endpoint.

[0073] According to various embodiments, the AI ​​agent service may be a service that provides personalized services based on personal data. When performing a call to use the AI ​​agent service in operation 5060, the first web3ai agent (100) may transmit personal data as a parameter. The first web3ai agent (100) may call the second web3ai agent (400) based on at least one parameter created based on the personal data (134).

[0074] In various embodiments, the first web3ai agent (100) and the second web3ai agent (400) can perform identity checks on each other (5050). For example, the first web3ai agent (100) can look up the identity information of the second web3ai agent (400) before calling the second web3ai agent (400). For example, the first web3ai agent (100) can look up the identity verification VC disclosed through the public certificate of the second web3ai contract (340). The first web3ai agent (100) can look up the identity information of the second web3ai agent (400) (e.g., information about the individual or company that owns and operates the second web3ai agent (400)) through the identity verification VC. The first web3ai agent (100) can call the second web3ai agent (400) only when it can trust the second web3ai agent (400) through the identity authentication VC (5060).

[0075] Additionally, when the second web3ai agent (400) receives a request for a call according to operation 5050, it may look up identity information to determine whether the first web3ai agent (100) is a user who has authority over the call (e.g., a user who has paid). For example, the second web3ai agent (400) may look up the identity authentication VC disclosed through the public certificate (312) of the first web3ai contract (310). Only if the result of the identity information lookup indicates that the first web3ai agent (100) has that authority, the second web3ai agent (400) may send a response according to operation 5060.

[0076] In various embodiments, the operations of FIG. 5 may be applied in the same way when the second web3ai agent (400) calls the first web3ai agent (100). When the first web3ai agent (100) receives a call from the second web3ai agent (400) according to the interface specified in the first web3ai contract (310) (ASI (314)) (5060), it can check whether the second web3ai agent (400) has authority over the call based on the identity authentication VC of the second web3ai agent (400) included in the second web3ai contract (340) of the second web3ai agent (400) registered in the blockchain network (300) (5050). If authority over the call is confirmed, the first web3ai agent (100) can execute the code corresponding to the call. The first web3ai agent (100) can generate a response to the call based on the result generated by the execution of the code and transmit the generated response to the second web3ai agent (400) (5070). The first web3ai agent (100) can execute the code corresponding to the call based on parameters created based on the personal data of the second web3ai agent (400).

[0077] In various embodiments, the first web3ai agent (100) may obtain the identity information by querying the identity authentication VC registered in the second web3ai contract corresponding to the second web3ai agent (400) / second web3ai-ID. Alternatively, the first web3ai agent (100) may transmit an identity information request to the second web3ai agent (400) and obtain the identity information from the identity authentication VP received from the second web3ai agent (400) in accordance with the request. The second web3ai agent (400) may generate an identity authentication VP based on the identity authentication VC stored in the second web3ai agent (400) and transmit the generated identity authentication VP to the first web3ai agent (100). The identity authentication VP may include a second web3ai-ID, an electronic signature generated based on a private key corresponding to the second web3ai-ID, an electronic signature of the creator of the identity authentication VC, and identity information requested by the first web3ai agent (100).

[0078] FIG. 6 is an example of a product recommendation service provided by a plurality of e-commerce service web3ai agents (410, 420, 430, 440) according to one embodiment. FIG. 7 is an example of an interface of a product recommendation service associated with an e-commerce service.

[0079] In various embodiments, the web3ai agent (100) can process product purchases on behalf of the web3ai-ID owner (user). The web3ai agent (100) can communicate with multiple e-commerce service web3ai agents (410, 420, 430, 440) to purchase products that are requested by the user or that are deemed necessary by the user. During this communication process, product search and purchase may be processed.

[0080] The web3ai agent (100) can search for web3ai agents operated by e-commerce service providers among the web3ai agents disclosed on the blockchain network (300).

[0081] A web3ai agent (100) can identify an AI agent service interface (ASI) (e.g., ASI (314) of FIG. 1) (hereinafter, product recommendation service) that provides product recommendations provided by e-commerce service web3ai agents (410, 420, 430, 440) through web3ai contracts (415, 425, 435, 445) (e.g., web3ai contract (310) of FIG. 1) corresponding to the e-commerce service web3ai agents (410, 420, 430, 440). Based on the interface of the identified product recommendation service, the web3ai agent (100) can receive a list of recommended products from the e-commerce service web3ai agents (410, 420, 430, 440).

[0082] Referring to FIG. 7, the interface (700) of the product recommendation service may include parameters of a search filter (710), associated personal data (720), and a purchase guide prompt (730). The interface (700) is an example of parameters created to recommend a cereal bar for a user's breakfast.

[0083] The web3ai agent (100) can automatically set a search filter (710) suitable for the user based on the personal data of the owner (user) of the web3ai-ID. For example, the search filter (710) can be set based on the user's previous order history, allergy information, preferred flavors, etc. included in the personal data (134).

[0084] The web3ai agent (100) can extract related personal data (720) necessary for product recommendation from personal data (134). For example, the web3ai agent (100) can construct related personal data (720) based on user data obtained through related applications (e.g., health, bio, e-commerce, SNS, etc.) and user health information (e.g., allergies, diseases, etc.).

[0085] Additionally, the web3ai agent (100) can create a purchase guide prompt (730) based on personal data (134). For example, the purchase guide prompt (730) can be created by the AI ​​agent module (124) based on information regarding the user's preferred taste and preferred food stored in the personal data (134).

[0086] The web3ai agent (100) can call e-commerce service web3ai agents (410, 420, 430, 440) based on a created search filter (710), associated personal data (720), and a purchase guide prompt (730). The web3ai agent (100) can call the e-commerce service web3ai agents (410, 420, 430, 440) by sending a product recommendation request containing the created parameters (search filter (710), associated personal data (720), and purchase guide prompt (730)) to the e-commerce service web3ai agents (410, 420, 430, 440) (e.g., operation 5060 of FIG. 5).

[0087] E-commerce service web3ai agents (410, 420, 430, 440) can select the most suitable products based on search filters (710), associated personal data (720), and purchase guide prompts (730) transmitted from the web3ai agent (100), and transmit a list of recommended products including the selected products to the web3ai agent (100) (e.g., operation 5070 of FIG. 5). For example, the service web3ai agents (410, 420, 430, 440) can execute an AI-based recommendation module (e.g., an AI model (138)) to select suitable products and generate a list of recommended products.

[0088] When the web3ai agent (100) receives a list of recommended products, it can compare the products included in the list and perform a purchase of the products. For example, the web3ai agent (100) can select N products from the list of recommended products and provide the list of N products to the user. And when the web3ai agent (100) receives a selection input from the list of N products, it can perform a purchase of the selected product.

[0089] As illustrated in FIGS. 6 and 7, the web3ai agent (100) can process the product purchase process by communicating directly with web3ai agents (410, 420, 430, 440) operated by e-commerce service providers, without using applications created by e-commerce service providers (appless process). Through the appless process, the web3ai-ID owner (user) can receive recommendations based on results searched through multiple e-commerce applications and conveniently purchase products without needing to search for products directly using multiple e-commerce applications.

[0090] In this way, web3ai agents can perform autonomous interactions through the blockchain-based web3ai-ID and web3ai contract protocols. Through this, tasks can be performed via the automated activities of web3ai agents without the user directly using the app.

[0091] The various embodiments of this document and the terms used therein are not intended to limit the technical features described in this document to specific embodiments, and should be understood to include various modifications, equivalents, or substitutions of said embodiments. In connection with the description of the drawings, similar reference numerals may be used for similar or related components. The singular form of a noun corresponding to an item may include one or more of said items unless the relevant context clearly indicates otherwise. In this document, phrases such as "A or B," "at least one of A and B," "at least one of A or B," "A, B or C," "at least one of A, B and C," and "at least one of A, B, or C" may each include any possible combination of items listed together in the corresponding phrase. Terms such as "first," "second," or "first" or "second" may be used simply to distinguish said components from other said components and do not limit said components in any other aspect (e.g., importance or order). Where any (e.g., 1st) component is referred to as “coupled” or “connected” to another (e.g., 2nd) component, with or without the terms “functionally” or “communicationly,” it means that said any component may be connected to said other component directly (e.g., via a wire), wirelessly, or through a third component.

[0092] As used in this document, the term "module" may include a unit implemented in hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit. A module may be a component formed as a whole, or a minimum unit of said component or a part thereof that performs one or more functions. For example, according to one embodiment, a module may be implemented in the form of an application-specific integrated circuit (ASIC).

[0093] Various embodiments of the present document may be implemented as software (e.g., program (136)) comprising one or more instructions stored in a storage medium (e.g., memory (120)) readable by a machine (e.g., web3ai agent (100)). For example, a processor (e.g., processor (110)) of the machine (e.g., web3ai agent (100)) may call at least one of the one or more instructions stored from the storage medium and execute it. This enables the machine to operate to perform at least one function according to the at least one called instruction. The one or more instructions may include code generated by a compiler or code that can be executed by an interpreter. The storage medium readable by the machine may be provided in the form of a non-transitory storage medium. Here, 'non-temporary' merely means that the storage medium is a tangible device and does not contain a signal (e.g., electromagnetic waves), and this term does not distinguish between cases where data is stored semi-permanently and cases where it is stored temporarily.

[0094] According to one embodiment, the method according to the various embodiments disclosed herein may be provided by being included in a computer program product. The computer program product may be traded between a seller and a buyer as a product. The computer program product may be distributed in the form of a device-readable storage medium (e.g., compact disc read-only memory (CD-ROM)) or an application store (e.g., Play Store). TM It can be distributed online (e.g., downloaded or uploaded) through ) or directly between two user devices (e.g., smartphones). In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily created on a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0095] According to various embodiments, each component (e.g., module or program) of the components described above may include a singular or multiple entities. According to various embodiments, one or more of the components or operations of the aforementioned components may be omitted, or one or more other components or operations may be added. Generally or additionally, multiple components (e.g., module or program) may be integrated into a single component. In this case, the integrated component may perform one or more functions of each of the components of the multiple components in the same or similar manner as those performed by the corresponding component among the multiple components prior to the integration. According to various embodiments, operations performed by the module, program, or other components may be executed sequentially, in parallel, iteratively, or heuristically, or one or more of the operations may be executed in a different order, omitted, or one or more other operations may be added.

Claims

Claim 1 In a first web3ai agent, a memory storing a first web3ai-ID corresponding to the first web3ai agent and a private key corresponding to the first web3ai-ID; and a communication interface that communicates with a blockchain network where a first web3ai contract corresponding to the first web3ai agent is registered. A first web3ai agent comprising at least one processor, wherein the at least one processor generates a pair of processor private keys and processor public keys, transmits the processor public key outside the at least one processor, executes code corresponding to at least one AI agent service specified in the first web3ai contract, obtains a code hash value corresponding to the executed code, obtains a state root value for personal data processed through the executed code, generates a first electronic signature for the code hash value and the state root value based on the processor private key, and is configured to transmit a registration request for the first electronic signature, which includes the first electronic signature and a second electronic signature generated based on the private key corresponding to the first web3ai-ID, to the blockchain network so that the first electronic signature is registered in the first web3ai contract. Claim 2 In claim 1, the first web3ai agent, wherein the at least one processor is configured to generate the first electronic signature at predetermined intervals and transmit a registration request for the generated first electronic signature to the blockchain network. Claim 3 A first web3ai agent according to claim 1, wherein the at least one processor is configured to generate the web3ai-ID and the private key and public key corresponding to the web3ai-ID, and to transmit to the blockchain network a request for the creation of a web3ai contract including an electronic signature generated based on the web3ai-ID and the private key, and the web3ai contract is created on the blockchain network corresponding to the web3ai-ID. Claim 4 A first web3ai agent according to claim 1, wherein the at least one processor is configured to perform communication with a second web3ai agent by querying the identity authentication VC of the second web3ai agent included in the second web3ai contract registered in the blockchain network, calling the second web3ai agent according to an interface specified in the second web3ai contract, and receiving a response to the call from the second web3ai agent. Claim 5 In claim 4, the memory stores personal data, and the at least one processor is configured to create at least one parameter included in the interface based on the personal data and to call the second web3ai agent based on the created at least one parameter, a first web3ai agent. Claim 6 A first web3ai agent according to claim 1, wherein the at least one processor is configured to, upon receiving a call from a second web3ai agent according to an interface specified in the first web3ai contract, determine whether the second web3ai agent has authority over the call based on identity information of the second web3ai agent, and execute the code corresponding to the call. Claim 7 In claim 6, the first web3ai agent, wherein the at least one processor is configured to execute the code corresponding to the call based on parameters created based on the personal data of the second web3ai agent. Claim 8 In claim 6, the at least one processor is configured to obtain the identity information by querying the identity authentication VC registered in the second web3ai contract corresponding to the second web3ai agent, or to transmit an identity information request to the second web3ai agent and obtain the identity information from the identity authentication VP received in accordance with the request, a first web3ai agent.

Images