Multi-agent security interaction method, equipment and medium

Abstract

The invention discloses a multi-agent security interaction method and device and a medium, belongs to the technical field of artificial intelligence and block chain crossing, and is used for solving the technical problems that an existing agent system has a single-point failure risk, direct communication between energy bodies is easily attacked by man-in-the-middle, and privacy data is easily excessively exposed in multi-party cooperation. The method comprises the following steps: carrying out hierarchical identity self-certification interaction control related to a block chain on an intelligent agent, and generating an identity authentication strategy of the intelligent agent; performing data credible interaction control related to the multi-stage data sandbox on the intelligent agent, and generating a data sandbox strategy; performing access control chain control based on block chain driving on the shared data of the intelligent agent to generate an access control chain strategy of the intelligent agent; performing data verification control related to alliance chain nodes on the cross-domain data of the intelligent agent to generate an alliance block chain cooperation strategy; and performing abnormal behavior monitoring control on the intelligent agent under network interaction to generate an abnormal behavior real-time monitoring strategy of the intelligent agent.

Description

Technical Field

[0001] This application relates to the intersection of artificial intelligence and blockchain, and in particular to a method, device and medium for secure interaction among multiple agents. Background Technology

[0002] An intelligent agent is a proxy capable of perceiving its environment and taking actions to achieve specific goals. It can be software, hardware, or a system, possessing autonomy, adaptability, and interactivity. Intelligent agents perceive changes in the environment (e.g., through sensors or data input), make judgments and decisions based on their learned knowledge and algorithms, and then execute actions to influence the environment or achieve predetermined goals. Intelligent agents are widely used in the field of artificial intelligence, commonly found in automated systems, robots, virtual assistants, and game characters. Their core strength lies in their ability to learn autonomously and continuously evolve to better complete tasks and adapt to complex environments.

[0003] Existing intelligent agent systems rely on centralized authentication servers, posing a single point of failure risk; furthermore, direct communication between intelligent agents is vulnerable to man-in-the-middle attacks. While blockchain can guarantee data immutability, the combination of traditional intelligent agents and blockchain lacks fine-grained access control and cross-domain collaboration mechanisms, leading to excessive exposure of private data during multi-party collaborations. Summary of the Invention

[0004] This application provides a method, device, and medium for secure interaction among multiple agents, which addresses the following technical problems: existing agent systems have single-point-of-failure risks, direct communication between agents is susceptible to man-in-the-middle attacks, and privacy data is easily overexposed in multi-party collaboration.

[0005] The embodiments of this application adopt the following technical solutions:

[0006] On one hand, embodiments of this application provide a multi-agent secure interaction method, including: performing layered identity self-authentication interaction control on the agents related to blockchain, generating an identity authentication strategy for the agents; wherein, the layering includes: an on-chain registration layer and a dynamic reputation evaluation layer; performing data trust interaction control on the agents under a multi-level data sandbox, generating a data sandbox strategy for the agents; performing access control chain control on the shared data of the agents based on blockchain, generating an access control chain strategy for the agents; performing data verification control on the cross-domain data of the agents related to consortium blockchain nodes, generating a consortium blockchain collaboration strategy for the agents; performing abnormal behavior monitoring control on the agents under network interaction, generating a real-time abnormal behavior monitoring strategy for the agents; and combining the identity authentication strategy, the data sandbox strategy, the access control chain strategy, the consortium blockchain collaboration strategy, and the real-time abnormal behavior monitoring strategy in a collaborative manner to generate a multi-agent secure interaction collaboration system.

[0007] This application's embodiments significantly improve the security of intelligent agent interactions and reduce security vulnerabilities and risks by comprehensively utilizing multiple security strategies. Data sandboxes and access control chains protect sensitive data between intelligent agents from unauthorized access and tampering. Furthermore, the combination of layered and collaborative strategies allows the intelligent agent collaboration system to adapt to different security needs and environmental changes. The integration of automated and intelligent security control mechanisms reduces manual intervention and improves system operating efficiency. Based on the technical characteristics of blockchain, a transparent auditing and traceability mechanism is provided, enhancing the system's credibility.

[0008] In one feasible implementation, the intelligent agent undergoes layered identity self-certification interaction control related to the blockchain to generate an identity authentication strategy for the intelligent agent. Specifically, this includes: when the intelligent agent first accesses the blockchain, generating a unique identity key using an asymmetric key, and storing the corresponding public key and metadata on the blockchain; using a zero-knowledge proof mechanism under the Sigma protocol, performing sensitive information identity authentication based on a three-step interaction under the Sigma protocol during the self-certification stage of the intelligent agent after storage, and accelerating the authentication calculation of the sensitive information based on pre-calculated and stored elliptic curve parameters to generate a first identity authentication strategy under the on-chain registration layer; dynamically adjusting the permission configuration related to identity authentication in the dynamic reputation evaluation layer to generate a second identity authentication strategy; and combining the first and second identity authentication strategies to obtain the intelligent agent's identity authentication strategy.

[0009] In one feasible implementation, a second identity authentication strategy is generated based on the dynamic permission adjustment configuration related to identity authentication in the dynamic reputation assessment layer. Specifically, this includes: calculating and processing the real-time reputation of the intelligent agent's own data based on multi-dimensional behavioral data stored on the blockchain in the dynamic reputation assessment layer, and obtaining a real-time reputation score for each intelligent agent through a preset smart contract; dynamically adjusting the three-level permissions corresponding to the intelligent agent based on the real-time reputation score to obtain a dynamic permission adjustment strategy; and using the dynamic permission adjustment strategy to perform identity self-authentication interaction control of the intelligent agent based on the dynamic reputation assessment layer, thereby generating the second identity authentication strategy.

[0010] In one feasible implementation, the agent is subjected to trusted data interaction control under a multi-level data sandbox, generating a data sandbox strategy for the agent. Specifically, this includes: performing isolated computation on the original data in the agent through a trusted execution environment, and encrypting it using a national cryptographic algorithm to obtain an encrypted feature vector; uploading the encrypted feature vector of the agent to blockchain storage; performing secondary encryption on the uploaded encrypted feature vector using preset differential privacy and homomorphic encryption techniques to obtain software-level multi-level encrypted data; performing hardware-level encrypted storage on absolutely shared data in the agent through a trusted execution environment, and generating a corresponding access control policy to obtain hardware-level encrypted data; wherein, the absolutely shared data is the data that the agent must share under blockchain processing; and generating the agent's data sandbox strategy based on the trusted interaction control method between the software-level multi-level encrypted data and the hardware-level encrypted data.

[0011] In one feasible implementation, the shared data of the intelligent agent is controlled by an access control chain based on blockchain to generate the access control chain policy of the intelligent agent. Specifically, this includes: training a model for the shared data in the intelligent agent under three-dimensional permissions through the access control chain driven by the blockchain to obtain a three-dimensional permission policy model; wherein the three-dimensional permissions include: role permissions, task permissions, and environment permissions; using the three-dimensional permission policy model, authorizing contract triggering control is applied to the shared data for data access, and verifying whether the data attributes of the access requester match the three-dimensional permission policy, and recording access logs; based on the three-dimensional permission policy of the three-dimensional permission policy model, the shared data of the intelligent agent is controlled by an access control chain driven by blockchain to generate the access control chain policy of the intelligent agent.

[0012] In one feasible implementation, data verification control of the agent's cross-domain data with respect to consortium blockchain nodes is performed to generate the agent's consortium blockchain collaboration strategy. Specifically, this includes: splitting the agent's cross-domain data to obtain a hash digest of the cross-domain data and the original cross-domain data; performing on-chain notarization processing on the hash digest and storing the original cross-domain data on a private blockchain; generating a consortium blockchain verification strategy based on the interactive comparison and verification of the hash digest; using the Byzantine fault-tolerant consensus algorithm in the consortium blockchain nodes to perform data operation maintenance control on some abnormal nodes, generating a consortium blockchain fault-tolerant strategy; and combining the consortium blockchain verification strategy and the consortium blockchain fault-tolerant strategy to obtain the agent's consortium blockchain collaboration strategy.

[0013] In one feasible implementation, the abnormal behavior monitoring and control of the intelligent agents under network interaction is performed to generate a real-time monitoring strategy for the abnormal behavior of the intelligent agents. Specifically, this includes: identifying each intelligent agent as graph node information and determining the communication relationships between the intelligent agents in network interaction as edge information; constructing a dynamic graph structure based on the time-series feature structure of the intelligent agent interaction mode; identifying abnormal communication features of the high-order communication relationship features in the dynamic graph structure through a preset graph neural network to determine abnormal communication behavior; triggering the multi-signature traceability mechanism and the abnormal event recording mechanism of the multi-supervisory node of the blockchain based on the abnormal communication behavior, and automatically executing the emergency response strategy through smart contracts; and generating a real-time monitoring strategy for the abnormal behavior of the intelligent agents based on the response execution results of the abnormal communication behavior.

[0014] In one feasible implementation, the identity authentication strategy, the data sandbox strategy, the access control chain strategy, the consortium blockchain collaboration strategy, and the real-time abnormal behavior monitoring strategy are combined in a collaborative manner to generate a multi-agent secure interaction and collaboration system. Specifically, this includes: performing integrated collaborative association processing between the identity authentication strategy, the data sandbox strategy, the access control chain strategy, the consortium blockchain collaboration strategy, and the real-time abnormal behavior monitoring strategy to obtain a multi-agent secure interaction module architecture; deploying the multi-agent secure interaction module architecture to obtain the multi-agent secure interaction and collaboration system; and using the multi-agent secure interaction and collaboration system to achieve trusted identity authentication and secure data sharing among the agents.

[0015] Secondly, embodiments of this application also provide a multi-agent secure interaction device, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to execute a multi-agent secure interaction method as described in any of the above embodiments.

[0016] Thirdly, embodiments of this application also provide a non-volatile computer storage medium, characterized in that the storage medium is a non-volatile computer-readable storage medium, the non-volatile computer-readable storage medium stores at least one program, each program includes instructions, and when the instructions are executed by a terminal, the terminal executes a multi-agent secure interaction method as described in any of the above embodiments.

[0017] This application provides a method, device, and medium for secure interaction among multiple agents. Compared with the prior art, the embodiments of this application have the following beneficial technical effects:

[0018] 1. Enhanced identity authentication security: Through layered identity self-authentication via on-chain registration and dynamic reputation assessment layers, the security of intelligent agent identity authentication is enhanced, reducing the risk of fraud and unauthorized access.

[0019] 2. Trusted Data Interaction: Trusted data interaction control of intelligent agents under multi-level data sandboxes helps to isolate sensitive data and prevent data leakage and unauthorized access.

[0020] 3. Blockchain-based access control: By utilizing a blockchain-driven access control chain, the transparency and immutability of data access are ensured, thereby improving the security of data in intelligent agents.

[0021] 4. Cross-domain data verification: Data verification control is performed on cross-domain data through consortium blockchain nodes to ensure the correctness and consistency of data across different intelligent agents.

[0022] 5. Abnormal Behavior Detection: By monitoring abnormal network interaction behavior of intelligent agents in real time, potential security threats can be quickly identified and responded to.

[0023] 6. Strategic Collaboration and Combination: Different security strategies are collaboratively combined to form a multi-agent security interaction and collaboration system, which improves the overall system's security and adaptability. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0025] Figure 1 A flowchart of a multi-agent secure interaction method provided in an embodiment of this application;

[0026] Figure 2 This is a schematic diagram of the structure of a multi-agent secure interaction device provided in an embodiment of this application. Detailed Implementation

[0027] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0028] This application provides a method for secure multi-agent interaction, such as... Figure 1 As shown, the multi-agent secure interaction method specifically includes steps S101-S106:

[0029] S101. Implement layered identity self-certification interaction control for the intelligent agent based on the blockchain, and generate the intelligent agent's identity authentication strategy. The layers include: an on-chain registration layer and a dynamic reputation assessment layer.

[0030] Specifically, when an intelligent agent first accesses the blockchain, it generates a unique identity key using an asymmetric key and stores the corresponding public key and metadata on the blockchain for evidence processing.

[0031] Furthermore, through the zero-knowledge identity proof mechanism under the Sigma protocol, the self-certification stage of the intelligent agent after notarization is performed to authenticate sensitive information based on the three-step interaction of the Sigma protocol. Based on the pre-calculated and stored elliptic curve parameters, the authentication of sensitive information is accelerated to generate the first identity authentication strategy under the on-chain registration layer.

[0032] Furthermore, based on the dynamic adjustment configuration of permissions under identity authentication in the dynamic reputation assessment layer, a second identity authentication strategy is generated.

[0033] As a feasible implementation, based on the multi-dimensional behavioral data stored on the blockchain in the dynamic reputation assessment layer, and through a pre-set smart contract, the real-time reputation of each intelligent agent is calculated and processed to obtain a real-time reputation score. Then, based on the real-time reputation score, the three levels of permissions corresponding to the intelligent agent are dynamically adjusted to obtain a dynamic permission adjustment strategy. Through the dynamic permission adjustment strategy, the intelligent agent is subjected to identity self-verification interaction control based on the dynamic reputation assessment layer, generating a second identity authentication strategy.

[0034] Furthermore, the first identity authentication strategy and the second identity authentication strategy are combined to obtain the identity authentication strategy of the intelligent agent.

[0035] In one embodiment, an "on-chain registration + agent self-certification" mechanism is adopted: when an agent first connects, it registers a unique identity key (KID) on the blockchain, and subsequent interactions use zero-knowledge proof (ZKP) to self-certify its identity, thus avoiding the exposure of sensitive information.

[0036] In one embodiment, at the on-chain registration layer, upon initial access, the agent generates a unique identity key (KID) using an asymmetric key and stores the public key and metadata on the blockchain to ensure identity uniqueness and immutability. Subsequent interactions employ a zero-knowledge proof (ZKP) protocol for self-verification. Through pre-computed elliptic curve parameters and optimization using the Groth16 algorithm, a single authentication process takes less than 50ms, avoiding sensitive information leakage while ensuring efficient verification.

[0037] In one embodiment, the dynamic reputation assessment layer uses multi-dimensional behavioral data (response latency, task completion rate, etc.) stored on the blockchain to calculate reputation scores in real time through smart contracts, thereby dynamically adjusting the three levels of permissions from VIP to restricted, effectively preventing malicious node attacks.

[0038] S102. Perform trusted data interaction control on the agent under the multi-level data sandbox and generate the agent's data sandbox strategy.

[0039] Specifically, it is also necessary to isolate and compute the raw data in the intelligent agent through a trusted execution environment, and then encrypt it based on the national cryptographic algorithm to obtain the encrypted feature vector.

[0040] Furthermore, the encrypted feature vector of the intelligent agent is uploaded to the blockchain storage.

[0041] Furthermore, by using preset differential privacy and homomorphic encryption technologies, the uploaded encrypted feature vector is subjected to secondary encryption processing to obtain software-level multi-level encrypted data.

[0042] Furthermore, through a trusted execution environment, the absolutely shared data within the intelligent agent is stored with hardware-level encryption, and a corresponding access control policy is generated, resulting in hardware-level encrypted data. Here, the absolutely shared data refers to the data that the intelligent agent must process through blockchain sharing.

[0043] Furthermore, a data sandbox strategy for intelligent agents is generated based on a trusted interactive control method that combines software-level multi-level encrypted data with hardware-level encrypted data.

[0044] In one embodiment, a multi-level data sandbox needs to be constructed to configure the agent's data sandbox strategy. Specifically, the agent preprocesses the raw data locally, isolates the computation process through a Trusted Execution Environment (TEE), and prevents the leakage of sensitive information. Only encrypted feature vectors are uploaded to the blockchain storage, reducing the amount of on-chain data while ensuring privacy. Simultaneously, differential privacy and homomorphic encryption technologies are used for secondary protection of the feature vectors, ensuring that even if the data is intercepted, the original information cannot be recovered. For raw data that must be shared, the system stores it in the TEE with hardware-level encryption and sets strict access control policies, allowing only authorized agents to access it through a secure channel, fundamentally eliminating the risk of data leakage.

[0045] S103. Perform blockchain-driven access control chain control on the shared data of the intelligent agent to generate the access control chain policy of the intelligent agent.

[0046] Specifically, through a blockchain-driven access control chain, shared data within intelligent agents is processed using a three-dimensional permission model to obtain a three-dimensional permission policy model. The three-dimensional permissions include: role permissions, task permissions, and environment permissions.

[0047] Furthermore, through a three-dimensional permission policy model, authorization contract triggering control is implemented for shared data access, and the data attributes of the access requester are verified to match the three-dimensional permission policy, and access logs are recorded.

[0048] Furthermore, based on the three-dimensional permission policy model, the three-dimensional permission policy controls the shared data of the intelligent agent through a blockchain-driven access control chain, generating the intelligent agent's access control chain policy.

[0049] In one embodiment, a blockchain-driven Access Control Chain (ACC) is used to configure access control policies for smart agents. First, ACC defines a three-dimensional permission policy model: "role-task-environment." Roles include data owners, analysts, auditors, etc.; tasks refer to specific operation types (e.g., read, modify, delete); and the environment defines contextual factors such as network conditions and time range. Data access triggers an on-chain authorization contract. The smart contract verifies in real-time whether the requester's attributes match the policy requirements and records a complete access log. This mechanism supports fine-grained, field-level permission control, such as allowing a smart agent to access only a specific portion of data within a specific time period, significantly improving the security and flexibility of data sharing. Furthermore, all permission change operations are accompanied by a blockchain timestamp, forming an immutable operation log. The smart contract has a built-in time window detector that automatically triggers subsequent cleanup processes when a permission revocation event is detected.

[0050] S104. Perform data verification and control on cross-domain data of the intelligent agent related to the consortium blockchain nodes, and generate the consortium blockchain collaboration strategy of the intelligent agent.

[0051] Specifically, the cross-domain data of the agent is split and processed to obtain a hash digest of the cross-domain data and the original cross-domain data.

[0052] Furthermore, the hash value digest is processed for blockchain-based on-chain notarization, and the original cross-domain data is stored on a private blockchain.

[0053] Furthermore, based on the interactive comparison and verification of hash value digests, a consortium blockchain verification strategy is generated.

[0054] Furthermore, by using the Byzantine fault-tolerant consensus algorithm in the consortium blockchain nodes, data operation maintenance control is implemented for some abnormal nodes, thereby generating a consortium blockchain fault-tolerant strategy.

[0055] Furthermore, by combining the consortium blockchain verification strategy with the consortium blockchain fault tolerance strategy, a consortium blockchain collaboration strategy for intelligent agents is obtained.

[0056] In one embodiment, a consortium blockchain collaboration strategy for intelligent agents is constructed using a consortium blockchain architecture. In cross-domain data sharing scenarios, the system only stores the hash digest of the data on the blockchain for notarization, while the original data is processed within the private blockchains of each participant. This avoids directly exposing sensitive information to the public network and ensures the privacy of data processing. To guarantee data integrity, the system employs an efficient hash verification mechanism: during each data interaction, the receiver recalculates the data hash and compares it with the notarized digest stored on the blockchain. If the verification fails, an anomaly alarm is immediately triggered. Simultaneously, the consortium blockchain nodes utilize a Byzantine fault-tolerant consensus algorithm to ensure the system continues to operate normally even if some nodes fail or are attacked, providing a reliable data exchange foundation for cross-organizational collaboration.

[0057] As a feasible implementation method, in a consortium blockchain architecture, permission states are synchronized through cross-chain communication protocols (such as Cosmos IBC) to ensure the consistency of permission data between the private and public blockchains. An eventual consistency model is adopted, coupled with an exponential backoff retry strategy, to ensure that permission revocation can still be achieved eventually even in the event of network partitioning. Simultaneously, a supporting abnormal communication monitoring system calculates the permission revocation success rate in real time and sets alarm thresholds (e.g., triggering an alarm if revocation delay > 90 seconds) to ensure system security and controllability.

[0058] S105. Monitor and control the abnormal behavior of the agent under network interaction, and generate a real-time monitoring strategy for the abnormal behavior of the agent.

[0059] Specifically, each agent must be identified as a graph node, and the communication relationships between agents in the network interaction must be identified as edge information. Based on the time-series feature structure of the agent interaction mode, a dynamic graph structure is constructed.

[0060] Furthermore, by using a pre-defined graph neural network, abnormal communication features are identified in the high-order communication relationship features of the dynamic graph structure, thereby determining abnormal communication behavior.

[0061] Furthermore, based on abnormal communication behavior, the blockchain's multi-signature tracing mechanism and the abnormal event recording mechanism of multiple regulatory nodes are triggered, and emergency response strategies are automatically executed through smart contracts.

[0062] Furthermore, based on the response execution results of abnormal communication behavior, a real-time monitoring strategy for abnormal behavior of the intelligent agent is generated.

[0063] In one embodiment, a real-time monitoring strategy for abnormal agent behavior is generated based on real-time monitoring of abnormal behavior between agents. This involves constructing a graph model of agent communication behavior using the system to analyze interaction patterns in the network in real time to identify potential threats. This model treats each agent as a graph node, its communication relationships as edges, and constructs a dynamic graph structure by combining time-series features. An anomaly detection module based on a graph neural network (GNN) extracts high-order relationship features through graph convolution operations, enabling accurate identification of abnormal communication patterns, such as data leakage attempts and malicious node injection. Once an anomaly is detected, the system immediately triggers a blockchain multi-signature traceability mechanism: multiple supervisory nodes jointly verify and record the anomaly, forming an immutable audit log. Simultaneously, emergency response strategies are automatically executed through smart contracts, such as isolating suspicious nodes and restricting data access permissions, forming a complete security closed loop from anomaly detection to handling, significantly improving the system's overall defense capabilities.

[0064] S106. Combine identity authentication strategy, data sandbox strategy, access control chain strategy, consortium blockchain collaboration strategy and real-time monitoring strategy for abnormal behavior to generate a multi-agent secure interaction and collaboration system.

[0065] Specifically, the identity authentication strategy, data sandbox strategy, access control chain strategy, consortium blockchain collaboration strategy, and real-time monitoring strategy for abnormal behavior need to be integrated and correlated to obtain a multi-agent security interaction module architecture.

[0066] Furthermore, the multi-agent safe interaction module architecture is then deployed to obtain a multi-agent safe interaction and collaboration system.

[0067] Furthermore, a multi-agent secure interaction and collaboration system is used to achieve trusted identity authentication and secure data sharing among agents.

[0068] In addition, embodiments of this application also provide a multi-agent secure interaction device, such as... Figure 2 As shown, the multi-agent secure interaction device 200 specifically includes:

[0069] At least one processor 201. And a memory 202 communicatively connected to the at least one processor 201. The memory 202 stores instructions executable by the at least one processor 201, enabling the at least one processor 201 to execute:

[0070] The system implements layered identity self-certification interaction control for intelligent agents based on blockchain technology, generating identity authentication strategies for intelligent agents; the layers include: an on-chain registration layer and a dynamic reputation assessment layer.

[0071] To perform trusted data interaction control on intelligent agents under a multi-level data sandbox, and generate data sandbox strategies for intelligent agents;

[0072] The shared data of intelligent agents is controlled by an access control chain based on blockchain, and the access control chain policy of the intelligent agents is generated.

[0073] Perform data verification and control on cross-domain data of intelligent agents related to consortium blockchain nodes, and generate consortium blockchain collaboration strategies for intelligent agents.

[0074] Detect and control abnormal behavior of intelligent agents under network interaction, and generate real-time monitoring strategies for abnormal behavior of intelligent agents.

[0075] By combining identity authentication strategies, data sandbox strategies, access control chain strategies, consortium blockchain collaboration strategies, and real-time monitoring strategies for abnormal behavior, a multi-agent secure interaction and collaboration system can be generated.

[0076] This application's embodiments significantly improve the security of intelligent agent interactions and reduce security vulnerabilities and risks by comprehensively utilizing multiple security strategies. Data sandboxes and access control chains protect sensitive data between intelligent agents from unauthorized access and tampering. Furthermore, the combination of layered and collaborative strategies allows the intelligent agent collaboration system to adapt to different security needs and environmental changes. The integration of automated and intelligent security control mechanisms reduces manual intervention and improves system operating efficiency. Based on the technical characteristics of blockchain, a transparent auditing and traceability mechanism is provided, enhancing the system's credibility.

[0077] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for devices and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0078] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0079] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0081] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0082] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0083] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The above description is merely an embodiment of this application and is not intended to limit this application. For those skilled in the art, various modifications and variations can be made to the embodiments of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the embodiments of this application should be included within the scope of the claims of this application.

Claims

1. A method for secure interaction among multiple agents, characterized in that, The method includes: The intelligent agent is subjected to layered identity self-certification interaction control related to the blockchain, and an identity authentication strategy for the intelligent agent is generated; wherein, the layering includes: an on-chain registration layer and a dynamic reputation evaluation layer; Perform trusted data interaction control on the intelligent agent under a multi-level data sandbox, and generate the data sandbox strategy of the intelligent agent; The shared data of the intelligent agent is controlled by an access control chain based on blockchain, and the access control chain policy of the intelligent agent is generated. The cross-domain data of the intelligent agent is subject to data verification and control of relevant consortium blockchain nodes, and a consortium blockchain collaboration strategy for the intelligent agent is generated. The abnormal behavior monitoring and control of the intelligent agent under network interaction is performed to generate a real-time monitoring strategy for the abnormal behavior of the intelligent agent. The identity authentication strategy, the data sandbox strategy, the access control chain strategy, the consortium blockchain collaboration strategy, and the real-time abnormal behavior monitoring strategy are combined to generate a multi-agent secure interaction and collaboration system.

2. The multi-agent secure interaction method according to claim 1, characterized in that, The system performs layered identity self-verification interaction control on the intelligent agent related to the blockchain, generating the intelligent agent's identity authentication strategy, specifically including: When the intelligent agent first accesses the blockchain, it generates a unique identity key using an asymmetric key and stores the corresponding public key and metadata on the blockchain for evidence processing. Through the zero-knowledge identity proof mechanism under the Sigma protocol, the self-certification stage of the intelligent agent after notarization is performed to authenticate sensitive information based on the three-step interaction of the Sigma protocol. Based on the pre-calculated and stored elliptic curve parameters, the authentication of the sensitive information is accelerated to generate the first identity authentication strategy under the on-chain registration layer. Based on the dynamic credit assessment layer, the permission configuration under identity authentication is dynamically adjusted to generate a second identity authentication strategy. The first identity authentication strategy and the second identity authentication strategy are combined to obtain the identity authentication strategy of the intelligent agent.

3. The multi-agent secure interaction method according to claim 2, characterized in that, Based on the dynamic credit assessment layer's dynamic permission adjustment configuration related to identity authentication, a second identity authentication strategy is generated, specifically including: Based on the multi-dimensional behavioral data stored on the blockchain in the dynamic reputation assessment layer, and through a preset smart contract, the real-time reputation score of each smart agent is calculated and processed in real time. Based on the real-time reputation score, the three-level permissions corresponding to the intelligent agent are dynamically adjusted to obtain a dynamic permission adjustment strategy. The permission dynamic adjustment strategy is used to perform identity self-authentication interaction control on the intelligent agent based on the dynamic reputation evaluation layer, thereby generating the second identity authentication strategy.

4. The multi-agent secure interaction method according to claim 1, characterized in that, The agent is subjected to trusted data interaction control under a multi-level data sandbox, and a data sandbox strategy for the agent is generated, specifically including: The original data in the intelligent agent is isolated and computed through a trusted execution environment, and encrypted using the national cryptographic algorithm to obtain an encrypted feature vector. The encrypted feature vector of the intelligent agent is uploaded to the blockchain storage; By using preset differential privacy and homomorphic encryption technology, the uploaded encrypted feature vector is subjected to secondary encryption processing to obtain software-level multi-level encrypted data. Through a trusted execution environment, the absolutely shared data in the intelligent agent is stored with hardware-level encryption, and a corresponding access control policy is generated to obtain hardware-level encrypted data; wherein, the absolutely shared data is the data that the intelligent agent must process under blockchain sharing. Based on the trusted interaction control method between the software-level multi-level encrypted data and the hardware-level encrypted data, the data sandbox strategy of the intelligent agent is generated.

5. The multi-agent secure interaction method according to claim 1, characterized in that, The shared data of the intelligent agent is subject to blockchain-driven access control chain control to generate the access control chain policy of the intelligent agent, specifically including: Through the access control chain driven by the blockchain, the shared data in the intelligent agent is subjected to model training processing under three-dimensional permissions to obtain a three-dimensional permission policy model; wherein, the three-dimensional permissions include: role permissions, task permissions and environment permissions. The three-dimensional permission policy model is used to control the authorization contract triggering of data access to the shared data, and to verify whether the data attributes of the access requester match the three-dimensional permission policy and record the access log. Based on the three-dimensional permission policy model, the shared data of the intelligent agent is controlled by a blockchain-driven access control chain, generating the access control chain policy of the intelligent agent.

6. The multi-agent secure interaction method according to claim 1, characterized in that, The cross-domain data of the intelligent agent is subject to data verification and control related to consortium blockchain nodes, and a consortium blockchain collaboration strategy for the intelligent agent is generated, specifically including: The cross-domain data of the intelligent agent is split and processed to obtain the hash digest of the cross-domain data and the original cross-domain data. The hash value digest is processed for blockchain-based on-chain notarization, and the original cross-domain data is stored on a private blockchain. Based on the interactive comparison and verification of the hash value digest, a consortium blockchain verification strategy is generated; By using the Byzantine fault-tolerant consensus algorithm in the consortium blockchain nodes, data operation maintenance control is performed on some abnormal nodes to generate a consortium blockchain fault-tolerant strategy. The consortium blockchain verification strategy and the consortium blockchain fault tolerance strategy are combined to obtain the consortium blockchain collaboration strategy of the agent.

7. A multi-agent secure interaction method according to claim 1, characterized in that, The abnormal behavior monitoring and control of the intelligent agent under network interaction is performed, and a real-time monitoring strategy for the abnormal behavior of the intelligent agent is generated, specifically including: Each agent is identified as graph node information, and the communication relationship between agents in the network interaction is identified as edge information. Based on the time series feature structure of the agent interaction mode, a dynamic graph structure is constructed. By using a pre-defined graph neural network, abnormal communication features are identified in the high-order communication relationship features of the dynamic graph structure, thereby determining abnormal communication behavior. Based on the abnormal communication behavior, the multi-signature tracing mechanism of the blockchain and the abnormal event recording mechanism of the multi-supervisory nodes are triggered, and the emergency response strategy is automatically executed through smart contracts. Based on the response execution results of the abnormal communication behavior, a real-time monitoring strategy for the abnormal behavior of the agent is generated.

8. A multi-agent secure interaction method according to claim 1, characterized in that, The identity authentication strategy, the data sandbox strategy, the access control chain strategy, the consortium blockchain collaboration strategy, and the real-time abnormal behavior monitoring strategy are combined in a collaborative manner to generate a multi-agent secure interaction and collaboration system, specifically including: The identity authentication strategy, the data sandbox strategy, the access control chain strategy, the consortium blockchain collaboration strategy, and the real-time abnormal behavior monitoring strategy are integrated and correlated to obtain a multi-agent security interaction module architecture. The multi-agent secure interaction module architecture is deployed and processed to obtain the multi-agent secure interaction and collaboration system. The multi-agent secure interaction and collaboration system enables trusted identity authentication and secure data sharing among the agents.

9. A multi-agent secure interaction device, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform a multi-agent secure interaction method according to any one of claims 1-8.

10. A non-volatile computer storage medium, characterized in that, The storage medium is a non-volatile computer-readable storage medium that stores at least one program, each program including instructions that, when executed by a terminal, cause the terminal to perform a multi-agent secure interaction method according to any one of claims 1-8.

Images