An Interpretable Artificial Intelligence Evidence Preservation Method and System Based on Blockchain

Abstract

The present invention relates to the fields of blockchain and artificial intelligence technologies, and particularly relates to an interpretable artificial intelligence evidence storage method and system based on blockchain. The method includes: establishing and polling a predictor that runs an artificial intelligence computing model to provide an explanation for the artificial intelligence result, establishing a reputation for each predictor, and storing data through a blockchain; interacting with the predictor through a smart contract on the blockchain and recording it on the ledger in an immutable manner; establishing a decentralized application for running an interpretable and trustworthy artificial intelligence application, while improving the delegated proof-of-stake consensus algorithm and using a Swarm storage platform to improve the consensus speed and data processing efficiency.

Description

Technical Field

[0001] The present invention relates to the technical fields of blockchain and artificial intelligence, and particularly relates to an explainable artificial intelligence evidence storage method and system based on blockchain. Background Art

[0002] Artificial intelligence (AI) currently has a revolutionary impact on many fields and industries. After long-term research and development, complex AI systems have been produced, and at the same time, complex external problems have begun to be faced, such as: when an artificial intelligence system correctly makes a decision on a data subset but performs poorly in the entire cluster, its susceptibility to bias increases; when some malicious actors attempt to manipulate data, resulting in incorrect decisions, an adversarial attack on the AI system will occur; the internal data representation and decision-making of the artificial intelligence system lack explanations.

[0003] Blockchain is a digital ledger that is publicly verifiable in a distributed environment and jointly maintained by a group of entities. It is composed of blocks containing a certain number of transactions connected in series. The generation time of a block is determined by all entities through an established consensus mechanism. Entities obtain certain rewards by competing to become the creator of a new block, and this reward is also the motivation for entities to participate in maintaining blockchain activities. In recent years, blockchain technology has developed rapidly and is more closely combined with other industry technologies, such as "blockchain + artificial intelligence", "blockchain + Internet of Things", "blockchain + finance", etc. Based on the continuous emergence of the integration of these technology industries, blockchain technology has been widely applied in fields such as security, digital forensics, healthcare, and financial institutions and achieved excellent results. Summary of the Invention

[0004] The present invention provides an explainable artificial intelligence evidence storage method and system based on blockchain. Its artificial intelligence technology using blockchain can provide decentralized learning, promote the trust and secure sharing of knowledge and decision-making results among a large number of autonomous agents. These agents can further contribute to, coordinate, and vote on decisions, and implement an open-source and accessible digital ledger that is distributed among AI agents through a peer-to-peer network, enabling AI agents to cooperate to execute consensus. Blockchain provides transparency and visibility of AI decisions for all participating AI agents on the network. Therefore, it is difficult for AI agents to change or reject decisions.

[0005] To achieve the above object, the present invention provides the following technical solution: An explainable artificial intelligence evidence storage method based on blockchain, which includes the following steps:

[0006] S1. Establish and poll a predictor that runs an artificial intelligence (AI) computing model to provide an explanation for the AI result, establish a reputation for each predictor, and store data through blockchain;

[0007] S2. Interact with the predictor through the smart contract SC on the blockchain and record it on the ledger in an immutable manner;

[0008] S3. Establish a decentralized application DApp for running interpretable and trustworthy AI applications.

[0009] Preferably, it further includes building an ecosystem of AI and XAI predictors, defining the node predictor ID as a consortium chain with the same prefix length, and its consortium chain members include predictors, data visitors, and a certificate authority CA.

[0010] Preferably, the consortium chain uses an improved delegated proof-of-stake DPOS consensus algorithm, and the predictor participates in the consensus as a full node. Twenty-one predictors are selected to participate in block production based on the predictor reputation vote.

[0011] Preferably, the AI and XAI predictors are registered through the CA, and each predictor has a reputation. When the artificial intelligence system calculates honestly, it is rewarded. The AI and XAI predictors directly or through other agnostic model interpretation systems provide explanations for their predictions, and reasonable explanations for the predictions contribute to the reliability of the decisions made.

[0012] Preferably, the Swarm storage platform is used in the blockchain to store the decision result data of the AI and XAI predictors in the Swarm platform; the information in the Swarm platform includes the data upload time of the evidence deposit, the node predictor ID, the current reputation status, the decision result, and the basic signature field information.

[0013] Preferably, an interpretable artificial intelligence evidence storage system based on blockchain, the DApp backend components mainly include four modules: AI access layer, AI layer, service layer and blockchain platform; wherein: the AI access layer realizes interfaces of various data transmission protocols; the Web3 interface is used for direct communication between the DApp and the blockchain platform; using the JSON-RPC API is conducive to transmitting data between the application program supporting the network and the blockchain network using the remote procedure call RPC; the AI layer is the main layer of the framework, and all data processing and knowledge discovery operations are executed here to generate trustworthy, collaborative and consistent decisions; the AI layer consists of predictors. According to the configuration received from the front-end Dapp, the AI and XAI predictors run on the original data and execute all priority operations or generate corresponding decisions by inputting the learning model and directly execute decisions on the already processed data; the service layer provides two types of support services: registration service and reputation service; the registration service can register and manage the participants in the system; the reputation service calculates and maintains the reputation of the AI and XAI predictors; the user selects a prediction node with good reputation through the Dapp, and the SC responsible for running the AI application on the predictor is set to automatically report the reputation score to the reputation SC when running each AI task; the blockchain platform includes: a blockchain network running the SC, and a decentralized memory for storing the results reported by the AI and XAI predictors; these SCs have different types, the registration SC supports decentralized registration and identification of AI / XAI predictors on the blockchain network; the AI work SC is responsible for collecting the final decision results from the aggregation SC and reporting the results to the front-end DApp; the aggregation SC is responsible for receiving the output from the predictors and reporting the reputation score of each predictor to the reputation SC.

[0014] The beneficial effects of the present invention are as follows: The artificial intelligence technology using blockchain can provide decentralized learning, promote the trust and secure sharing of knowledge and decision-making results among a large number of autonomous agents, and these agents can further contribute to, coordinate and vote on decisions, and implement an open-source and accessible digital ledger, which is distributed among AI agents through a peer-to-peer network, enabling AI agents to collaboratively execute consensus. Blockchain provides transparency and visibility of AI decisions for all participating AI agents on the network, so it is difficult for AI agents to change or reject decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0016] Figure 1 Schematic diagram of the process of the present invention;

[0017] Figure 2 Schematic diagram of the blockchain consensus process in the present invention;

[0018] Figure 3 Schematic diagram of the storage evidence data structure in the present invention;

[0019] Figure 4 Schematic diagram of the basic operation process of the backend service and platform framework of the present invention. Specific embodiments

[0020] Next, in combination with the accompanying drawings of the present invention, the technical solutions of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0021] According to Figure 1 , Figure 2 , Figure 3 , Figure 4 As shown, an explainable artificial intelligence evidence storage method based on blockchain includes the following steps:

[0022] S1. Establish and poll the predictor that runs the AI calculation model, provide an explanation for the AI result, establish a reputation for each predictor, and store the data through the blockchain.

[0023] Specifically, in this embodiment, an ecosystem of AI and XAI (Explainable AI) predictors is built, and the node predictor ID is defined as a consortium chain with the same prefix length.

[0024] An improved DPOS consensus algorithm is used in the consortium chain, and the predictor participates in the consensus as a full node. The consensus process is as Figure 2 shown. According to the predictor reputation voting, 21 predictors are selected to participate in block production, and 1 predictor node is the main node, that is: predictor i, and the remaining 20 predictors are predictor j nodes for block packaging. For example: 21 block producers are voted out, and the characteristic that the block producer is 100% online ensures that the consensus node will surely know a transaction within 1.5s, enabling the blockchain network to reach consistency. Therefore, only after the whole network node consensus passes, the blockchain can be effective.

[0025] The AI and XAI predictors directly or through other agnostic model interpretation systems provide explanations for their predictions. Predictions with reasonable explanations contribute to the reliability of the decisions made. Before use, these predictors must be registered with the CA.

[0026] S2. Interact with the predictor through the SC on the blockchain and record it on the ledger in an immutable manner.

[0027] Specifically, the AI and XAI predictors perform calculations and interact with the blockchain SC, recording the execution results and decisions in the immutable ledger of the blockchain.

[0028] Furthermore, the AI predictor stores the evidence of the decision results in the Swarm platform. As Figure 3 shown, the information in the Swarm platform includes the data upload time of the evidence storage, the node predictor ID, the current reputation status, the decision result, and the basic information of the signature field. These decision results include the type of decision, the value of the evaluation index, the confidence value, the explanation about the decision, and the type of explanation.

[0029] The reputation of the predictor is gradually established and changed by using the SC of the blockchain.

[0030] S3. Establish a decentralized application (DApp) for running interpretable and trustworthy AI applications.

[0031] A system applying the above method, its DApp backend components mainly include four modules: the AI access layer, the AI layer, the service layer, and the blockchain platform. The basic operation process of the framework is as Figure 4 shown.

[0032] The AI access layer implements the interfaces of various data transmission protocols. The Web3 interface is used for direct communication between the DApp and the blockchain platform. It uses the JSON-RPC API, which is beneficial for transmitting data between the application supporting the network and the blockchain network by using remote procedure call (RPC).

[0033] The AI layer is the main layer of the framework. All data processing and knowledge discovery operations are executed here to generate trustworthy, collaborative, and consistent decisions. This layer consists of predictors. According to the configuration received from the front-end Dapp, the AI and XAI predictors run on the original data and execute all priority operations or generate corresponding decisions by inputting the learning model and directly execute decisions on the already processed data.

[0034] The service layer provides two types of support services: registration service and reputation service. The registration service can register and manage the participants in the system. The reputation service calculates and maintains the reputation of AI and XAI predictors. Before running the AI application, the user selects a predictor with a good reputation through the Dapp. The SC responsible for running the AI application on the predictor can be set to automatically report the reputation score to the reputation SC when each AI task is run. If the AI and XAI predictors cannot report decision results that match those of most predictors, they will receive negative values and be penalized in terms of payment rewards.

[0035] The blockchain platform mainly includes: a blockchain network running SCs, and a decentralized memory for storing the results reported by AI and XAI predictors. These SCs have different types. The registration SC supports decentralized registration and identification of AI / XAI predictors on the blockchain network. The AI working SC is responsible for collecting the final decision results from the aggregation SC and reporting the results to the front-end DApp. The aggregation SC is responsible for receiving the outputs from the predictors and reporting the reputation score of each predictor to the reputation SC. Dishonest predictors will be penalized in terms of payment rewards and reputation scores.

[0036] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An explainable artificial intelligence evidence storage method based on blockchain, characterized in that , including the following steps: S1. Establish and poll the predictors that run the artificial intelligence (AI) computing model, provide explanations for the AI results, establish reputations for each predictor, and store data through the blockchain; S2. Interact with the predictors through the smart contract (SC) on the blockchain and record it on the ledger in an immutable manner; S3. Establish a decentralized application (DApp) for running an interpretable and trustworthy AI application; It also includes building an ecosystem of AI and XAI predictors. Define the node predictor ID as a consortium chain with the same prefix length. The members of its consortium chain include predictors, data visitors, and a certificate authority (CA); The consortium chain uses an improved delegated proof of stake (DPOS) consensus algorithm. The predictors participate in the consensus as full nodes, and 21 predictors are selected to participate in block production according to the predictor reputation voting; The AI and XAI predictors are registered through the CA. Each predictor has a reputation. When the artificial intelligence system calculates honestly, it is rewarded. The AI and XAI predictors directly or through other agnostic model interpretation systems provide explanations for their predictions. Predictions with reasonable explanations contribute to the reliability of the decisions made; The DApp backend components mainly include four modules: the AI access layer, the AI layer, the service layer, and the blockchain platform; among them: The AI access layer implements the interfaces of various data transmission protocols; the Web3 interface is used for direct communication between the DApp and the blockchain platform; the JSON-RPC API is used, which is conducive to using remote procedure call (RPC) to transmit data between the application supporting the network and the blockchain network; The AI layer is the main layer of the framework. All data processing and knowledge discovery operations are performed here to generate trustworthy, collaborative, and consistent decisions; the AI layer consists of predictors. According to the configuration received from the front-end Dapp, the AI and XAI predictors run on the original data and execute all priority operations or generate corresponding decisions by inputting the learning model and directly execute decisions on the already processed data; The service layer provides two types of support services: registration service and reputation service; the registration service can register and manage the participants in the system; the reputation service calculates and maintains the reputations of the AI and XAI predictors; users select predictor nodes with good reputations through the Dapp, and the SC responsible for running the AI application on the predictor is set to automatically report the reputation score to the reputation SC when running each AI task; The blockchain platform includes: a blockchain network running the SC, and a decentralized memory to store the results reported by the AI and XAI predictors; these SCs have different types. The registration SC supports decentralized registration and identification of AI / XAI predictors on the blockchain network; the AI work SC is responsible for collecting the final decision results from the aggregation SC and reporting the results to the front-end DApp; the aggregation SC is responsible for receiving the outputs from the predictors and reporting the reputation score of each predictor to the reputation SC.

2. The method for storing evidence of interpretable artificial intelligence based on blockchain according to claim 1, wherein: In the blockchain, the Swarm storage platform is used to store the decision result data of AI and XAI predictors in the Swarm platform; the information in the Swarm platform includes the data upload time of the evidence deposit, the node predictor ID, the current reputation status, the decision result, and the basic information of the signature field.

Images