An agent memory permission control method, device, equipment and program product

Abstract

Embodiments of the present application provide a memory permission control method, device and equipment of an agent, and a program product, wherein the memory permission control method of the agent is applied to a processor of the agent, and comprises: loading a pre-configured permission configuration table when it is detected that a third-party skill module of the agent is started; the permission configuration table stores access permissions of the third-party skill module to memory regions of different security levels; in a running process of the third-party skill module, whether there is an access behavior that exceeds the access permissions is determined based on the permission configuration table and an access behavior of the third-party skill module to the memory regions; and if there is, the access is refused. The memory permission control method of the agent can control the access behavior of the third-party skill module to the memory, and can improve the reliability of the agent.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, specifically to a method, apparatus, device, and program product for controlling memory permissions of an intelligent agent. Background Technology

[0002] In recent years, Large Language Models (LLMs) have driven AI (Artificial Intelligence) to shift from passive response to autonomous execution, with AI Agents becoming the core application form. Using the LLM as its "brain," it constructs a closed loop of perception-planning-action-reflection, enabling it to autonomously decompose tasks, invoke tools, remember context, and dynamically adjust.

[0003] However, AI agents use a full-authority execution strategy by default. This strategy allows users to execute system commands without granting authorization each time, but it also poses serious security risks. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method, apparatus, device, and program product for controlling memory permissions of an intelligent agent, so as to reduce the security risks of the intelligent agent.

[0005] In a first aspect, embodiments of the present invention provide a memory access control method for an intelligent agent, applied to the processor of the intelligent agent, comprising: When the activation of a third-party skill module of the intelligent agent is detected, a pre-configured permission configuration table is loaded; the permission configuration table stores the access permissions of the third-party skill module to memory regions with different security levels; During the operation of the third-party skill module, based on the permission configuration table and the access behavior of the third-party skill module to the memory area, it is determined whether there is any access behavior that exceeds its access permissions; if so, access is denied.

[0006] In a second aspect, embodiments of the present invention provide a memory access control device for an intelligent agent, applied to the processor of the intelligent agent, comprising: A loading module is used to load a pre-configured permission configuration table when the activation of a third-party skill module of the intelligent agent is detected; the permission configuration table stores the access permissions of the third-party skill module to memory regions of different security levels; The access control module is used to determine, during the operation of the third-party skill module, whether there is any access behavior that exceeds its access permissions based on the access configuration table and the access behavior of the third-party skill module to the memory area; if so, the access is denied.

[0007] Thirdly, embodiments of the present invention provide a computer program product, including a computer program that, when executed, implements the memory access control method for an intelligent agent as described in the first aspect.

[0008] Fourthly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a program, and the processor calls the program stored in the memory to execute the memory access control method for an intelligent agent as described in the first aspect.

[0009] Fifthly, embodiments of the present invention provide a storage medium storing a program that, when executed, implements the memory access control method for an intelligent agent as described in the first aspect.

[0010] This invention provides a memory access control method for an intelligent agent, applied to the processor of the intelligent agent, comprising: when a third-party skill module of the intelligent agent is detected to be started, loading a pre-configured access permission table; the access permission table stores the access permissions of the third-party skill module to memory regions of different security levels; during the operation of the third-party skill module, determining whether there is any access behavior exceeding its access permission based on the access permission table and the access behavior of the third-party skill module to the memory region; if so, denying access.

[0011] The memory access control method for an intelligent agent provided in this embodiment of the invention is applied to the processor of the intelligent agent. When a third-party skill module of the intelligent agent is detected to be starting, a pre-configured access permission table is loaded. This access permission table stores the access permissions of the third-party skill module to memory regions of different security levels. In other words, the memory access control method for the intelligent agent provided in this embodiment of the invention divides the memory of the intelligent agent into different security levels, and the access permission table stores the access permissions of the third-party skill module to the pre-defined memory regions of different security levels. Based on the access permission configuration table, it can be determined whether the third-party skill module engages in access behavior exceeding its access permissions during operation. If such access behavior exists, access to the memory region exceeding its access permissions is denied to the third-party skill module. This allows for access control of the memory access behavior of third-party skill modules within the intelligent agent, ensuring that the third-party skill module can only access the memory region corresponding to its access permissions, improving the security of data in memory, thereby reducing the security risks of the intelligent agent and improving its reliability. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0013] Figure 1 A flowchart illustrating the memory access control method for an intelligent agent provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the memory permission control device for an intelligent agent provided in an embodiment of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] In recent years, Large Language Models (LLMs) have driven AI to shift from passive response to autonomous execution, with AI Agents becoming the core application form. Using the LLM as its "brain," it constructs a closed loop of perception-planning-action-reflection, enabling it to autonomously decompose tasks, invoke tools, remember context, and dynamically adjust.

[0016] From early rule-based AI agents to today's LLM-driven autonomous agents, the technology has made leaps and bounds, no longer limited to fixed scripts, but processing complex and open tasks through thought chains, tool calls, and long-term memory.

[0017] In related technologies, an AI agent, centered on local deployment, autonomous control, and cross-messaging platform compatibility, enables efficient operation on user hardware, supports third-party skills, and achieves highly scalable autonomy, becoming an important direction for the next generation of personal and collaborative AI agents. However, as a local, all-encompassing AI agent, this agent defaults to a full-authority execution policy, allowing users to execute system commands without requiring individual authorization. While this design enhances the user experience, it also introduces serious security risks.

[0018] The AI ​​agent has the following major security vulnerabilities: 1. Due to hallucinations or loss of security commands caused by context compression, important user data may be accidentally deleted.

[0019] 2. Third-party skill modules use default full permissions to access unauthorized system resources and sensitive user data.

[0020] 3. Software-based access control schemes are easily bypassed and rely on AI to correctly execute security commands. Once the AI ​​loses access to security commands, the security commands become ineffective.

[0021] To address the above problems, this invention provides a memory access control method for an intelligent agent, applied to the processor of the intelligent agent. Please refer to [link / reference needed]. Figure 1 , Figure 1 This is a flowchart illustrating the memory permission control method for an intelligent agent provided in an embodiment of the present invention.

[0022] The processor is a processor that supports MPK (Memory Protection Keys). A processor that supports memory can divide memory into multiple regions and control access permissions between different memory regions through hardware-level permissions.

[0023] like Figure 1 As shown, the memory access control method for the intelligent agent includes: Step S101: When the third-party skill module of the intelligent agent is detected to be started, load the pre-configured permission configuration table.

[0024] Third-party skill modules are external extension components of an agent that are independent of the agent's core program. They are developed by third parties and include tool code, interface configuration, and execution logic. They are used to extend additional capabilities such as search and drawing. They run in an isolated sandbox and are modules with restricted access to highly sensitive memory data.

[0025] The permission configuration table stores the access permissions of the third-party skill module to memory regions with different security levels.

[0026] Step S102: During the operation of the third-party skill module, based on the permission configuration table and the access behavior of the third-party skill module to the memory area, determine whether there is any access behavior that exceeds its access permissions; if so, deny access.

[0027] The memory access control method for an intelligent agent provided in this embodiment of the invention is applied to the processor of the intelligent agent. When a third-party skill module of the intelligent agent is detected to be starting, a pre-configured access permission table is loaded. This access permission table stores the access permissions of the third-party skill module to memory regions of different security levels. In other words, the memory access control method for the intelligent agent provided in this embodiment of the invention divides the memory of the intelligent agent into different security levels, and the access permission table stores the access permissions of the third-party skill module to the pre-defined memory regions of different security levels. Based on the access permission configuration table, it can be determined whether the third-party skill module engages in access behavior exceeding its access permissions during operation. If such access behavior exists, access to the memory region exceeding its access permissions is denied to the third-party skill module. This allows for access control of the memory access behavior of third-party skill modules within the intelligent agent, ensuring that the third-party skill module can only access the memory region corresponding to its access permissions, improving the security of data in memory, thereby reducing the security risks of the intelligent agent and improving its reliability.

[0028] In one embodiment, the intelligent agent further includes a core program module; the permission configuration table also stores the access permissions of the core program module to memory regions with different security levels.

[0029] In this embodiment of the invention, the full lifecycle stages of the intelligent agent include: system initialization stage, user configuration stage, and operation control stage.

[0030] The user configuration phase is where users customize the parameters of the intelligent agent. In this phase, users select models, prompts, available tools, memory rules, permissions, workflows, output formats, etc., and generate a unique configuration file.

[0031] The system initialization phase mainly involves loading the agent's configuration, loading the agent's weights, initializing the memory, registering tools / functions, initializing state variables, loading the environment context, and verifying permissions. This completes all the pre-running preparations and is the startup phase where the agent is transformed from being unready to being operational.

[0032] The operation control phase is the stage where the agent enters a continuous work cycle after initialization. It involves receiving user input / environmental awareness, planning and thinking, invoking tools to execute actions, updating memory and state, and providing feedback results. This cyclical scheduling of agent reasoning, tool invocation, and state transitions is the core working stage for the agent's task processing. In the memory permission control method for the agent provided in this embodiment of the invention, steps S101 and S102 are both steps executed during the operation control phase. That is, steps S101 and S102 are essentially steps that control the access permissions of third-party skill modules to memory through a permission configuration table while the agent is working.

[0033] Based on the above discussion, it can be seen that before the operation control phase, there is a system initialization phase. Therefore, before loading the pre-configured permission configuration table when the third-party skill module of the intelligent agent is detected to be started, the memory permission control method of the intelligent agent in the system initialization phase further includes: forming a permission configuration table. The steps for forming the permission configuration table include: The security level of each memory region is determined based on the stored data objects in each memory region; Based on the security level of the memory region, determine the access permissions of the core program module and the third-party skill module under different access behaviors to the memory region; The permission configuration table is formed by storing memory regions with different security levels and their corresponding stored data objects, as well as the access permissions of core program modules and third-party skill modules to memory regions with different security levels.

[0034] Based on the stored data objects in each memory region, the security level of each memory region is determined. Then, based on the security level of the memory region, the access permissions of the core program module and the third-party skill module to that security level are determined. Essentially, determining the access permissions of the core program module and the third-party skill module to the memory region is based on the stored data objects. This allows different access permissions to be assigned to the core program module and the third-party skill module for the stored data objects in different memory regions, preventing the third-party skill module from using the default full-privilege mechanism to access unauthorized system resources and sensitive user data. This reduces the risk of third-party skill modules accessing system resources and sensitive user data in the intelligent agent, thus improving the security of the intelligent agent.

[0035] The security level, stored data objects, and access permissions of core program modules and third-party skill modules for each memory region are stored in the permission configuration table. This permission configuration table can be directly used in the subsequent operation control phase to restrict the access permissions of third-party skill modules and core program modules.

[0036] The stored data objects may include, for example, local critical and sensitive data, external skill data, and general user data; Local critical and sensitive data could include core programs, private keys, credentials, and important data tagged with user information. External skill data could include third-party skill code data and runtime data.

[0037] The core program is the underlying operational logic of the intelligent agent, serving as the foundational code and inference control logic driving the agent's autonomous planning, thinking, tool scheduling, memory management, and task decomposition. It does not contain business skills. The private key is the agent's exclusive encryption key and identity private key, used for local encrypted signing and authentication. It can be a symmetric or asymmetric private key. The credential is an authorization ticket for external access to third-party services, platform interfaces, and tools, used for network authentication, and is distinct from the local private key (the private key is encrypted internally, and the credential is used for external authorization).

[0038] User-tagged critical data refers to data manually marked by the user and labeled as private / high-value exclusive data. The sensitivity level is actively defined by the user, and the data is stored separately and isolated by the agent. User-defined general data refers to routine interaction content without special user markings and with low sensitivity levels. It is the most basic contextual data for agent conversations.

[0039] Third-party skill code data consists of the runtime code and built-in static data of external plugins, tools, and custom extended skills loaded by the agent. It is not the native core program and is written by external developers.

[0040] The runtime data are temporary intermediate calculation results dynamically generated during the execution of a single task by the agent, and only serve the current round of inference / task.

[0041] In the foregoing discussion, the security level of the memory region is determined based on the stored data objects in that memory region. In one embodiment, determining the security level of each memory region based on the stored data objects in each memory region includes: When a memory area stores local critical sensitive data (such as core programs, private keys, credentials, and important data marked by users), the security level of that memory area is determined to be the first security level, and that memory area is designated as the core sensitive area; when a memory area stores ordinary user data, the security level of that memory area is determined to be the second security level, and that memory area is designated as the ordinary user area; when a memory area stores external skill data (such as third-party skill code data and runtime data), the security level of that memory area is determined to be the third security level, and that memory area is designated as the skill runtime area. In terms of security level, the first security level is higher than the second security level, and the second security level is higher than the third security level.

[0042] Important data tagged by users can be, for example, tagged during the user configuration phase.

[0043] It can be seen that the security level of a memory region can be determined based on the stored data objects in the memory region. The security level of the stored data objects in the memory region reflects the security level of the memory region. Therefore, when the security level of a memory region is determined based on the stored data objects in the memory region, it is convenient to determine the access permissions of the core program module and the third-party skill module to the memory region according to the security level of the memory region.

[0044] In one embodiment, when the access behavior includes read, write, and delete, determining the access permissions of the core program module and the third-party skill module under different access behaviors of the memory region based on the security level of the memory region includes: When the security level of the memory region is the first security level, the core program module is determined to have access to the memory region, while the third-party skill module is determined to have access to the memory region but not. When the security level of the memory region is the second security level, the access permission of the core program module to the memory region is determined to be accessible, and the access permission of the third-party skill module to the memory region is readable, but not writeable or deletable; When the security level of the memory region is the third security level, the access rights of the core program module to the memory region are determined to be accessible, and the access rights of the third-party skill module to the memory region are also determined to be accessible.

[0045] As can be seen here, firstly, the security level of a memory region is determined based on the data objects stored in that region. The access permissions of the core program module and third-party skill modules to that memory region are determined based on their specific access behaviors within that security level. Therefore, in essence, the access permissions of the core program module and third-party skill modules to that memory region are also determined based on the data objects stored in that region. For example, if the core program is stored in a core sensitive area with a security level of Level 1, and this core sensitive area can be accessed by a third-party skill module, then in a malicious scenario, the core program could be tampered with by the third-party skill module. Similarly, if important data marked by the user is also stored in the core sensitive area, and this core sensitive area can be accessed by a third-party skill module, then the important data marked by the user could also be accessed by the third-party skill module. Therefore, the core sensitive area with a security level of Level 1 can only be accessed by the core program module and cannot be accessed by third-party skill modules.

[0046] Therefore, by determining the access permissions of the core program module and the third-party skill module under different access behaviors of the memory region based on the security level of the memory region, it is possible to directly prevent the core program module and the third-party skill module from accessing memory regions with security levels beyond their own access capabilities. This also prevents third-party skills from using default full permissions to access unauthorized system resources and sensitive user data, thereby improving the security of data objects stored in high-security memory regions.

[0047] In other embodiments, for example, when the memory area stores important data marked by the user, the security level of the memory area is determined to be the fourth security level, which is used as the user sensitive data area. The fourth security level is greater than the first security level. The user sensitive data area is set to be unwriteable and unmodifiable by the core program module and unaccessible by third-party skill modules. This can prevent the intelligent agent from accidentally deleting important user data due to illusions or context compression that may cause the loss of security instructions.

[0048] The security instructions refer to the system prompts written into the intelligent agent system. These are the highest-priority, unalterable underlying constraint rules used to restrict the intelligent agent's autonomous decision-making, tool invocation, content generation, and data reading and writing behaviors, thus preventing security risks from the software level.

[0049] In one embodiment, the permission configuration table may be as shown in Table 1 below: memory area Security level Stored data objects Access permissions Core Sensitive Area Level 1 security Core program, private key, credentials, and important user-tagged data Only the core program modules are accessible; third-party skill modules are not. Regular User Area Second security level User General Data Third-party skill modules are readable but not writable / deletable; core program modules are accessible. Skill Operation Area Third security level Third-party skill code data and runtime data Both the core program module and the third-party skill module are accessible. Table 1 Based on the above discussion, during the operation control phase, determining whether there is any access behavior exceeding the access permissions of the third-party skill module, based on the permission configuration table and the third-party skill module's access behavior to the memory region, includes: When the third-party skill module accesses the core sensitive area during operation, it is determined that the third-party skill module has access behavior exceeding its access permissions during operation. When the third-party skill module accesses the ordinary user area during operation, it is determined whether the access behavior is the third-party skill module reading data in the ordinary user area. If yes, it is determined that the third-party skill module does not have access behavior exceeding its access permissions during operation; otherwise, it is determined that the third-party skill module has access behavior exceeding its access permissions during operation. When the third-party skill module accesses the skill operation area during operation, it is determined that the third-party skill module does not have any access behavior exceeding its access permissions during operation.

[0050] Since the intelligent agent also includes a core program module, the memory permission control method for the intelligent agent during the runtime control phase further includes: When the core program module of the intelligent agent is detected to be started, a pre-configured permission configuration table is loaded; During the operation of the core program module, based on the permission configuration table and the core program module's access behavior to the memory region, it is determined whether there is any access behavior that exceeds its access permissions; if so, access is denied.

[0051] Optionally, during the execution of the core program module, determining whether there is any access behavior exceeding its access permissions based on the permission configuration table and the core program module's access behavior to the memory region includes: When the core program module accesses the core sensitive area, the ordinary user area, and the skill operation area during operation, there is no access behavior that exceeds its access permissions.

[0052] Based on the aforementioned permission configuration table, it is determined whether third-party skill modules and core program modules engage in access behavior exceeding their access permissions during operation. When such access behavior is detected, access to memory areas beyond their access permissions is denied. This allows for permission control over memory access behavior by third-party skill modules and core program modules within the intelligent agent, ensuring they can only access memory areas corresponding to their access permissions. This improves data security in memory, reduces security risks for the intelligent agent, and enhances its reliability.

[0053] In one embodiment, when access is denied, the memory permission control method of the intelligent agent further includes: storing the event corresponding to the access denial to a security log.

[0054] Storing access denial events in the security log will alert users when unauthorized access occurs in third-party skill modules or core program modules, allowing them to make timely adjustments and avoid losses.

[0055] Based on the same inventive concept, embodiments of the present invention also provide a memory access control device for an intelligent agent, applied to the processor of the intelligent agent. Please refer to [link / reference needed]. Figure 2 , Figure 2 This is a schematic diagram of the memory permission control device for an intelligent agent provided in an embodiment of the present invention.

[0056] like Figure 2As shown, the memory access control device for the intelligent agent includes: The loading module 201 is used to load a pre-configured permission configuration table when the startup of a third-party skill module of the intelligent agent is detected; the permission configuration table stores the access permissions of the third-party skill module to memory regions of different security levels; The access control module 202 is used to determine, during the operation of the third-party skill module, whether there is any access behavior that exceeds its access permissions based on the access configuration table and the access behavior of the third-party skill module to the memory area; if so, the access is denied.

[0057] The intelligent agent memory permission control device provided in this embodiment of the invention is applied to the processor of the intelligent agent. When a third-party skill module of the intelligent agent is detected to be running, a pre-configured permission configuration table is loaded. The permission configuration table includes the access permissions of the third-party skill module to memory regions with different pre-defined security levels. In other words, the intelligent agent memory permission control device provided in this embodiment of the invention divides the memory of the intelligent agent into different security levels, and the permission configuration table stores the access permissions of the third-party skill module to the pre-defined memory regions with different security levels. Based on the permission configuration table, it can be determined whether the third-party skill module engages in access behavior exceeding its access permissions during operation. If such access behavior is found, the third-party skill module is denied access to the memory region exceeding its access permissions. This allows for permission control of the third-party skill module's memory access behavior within the intelligent agent, ensuring that the third-party skill module can only access the memory region corresponding to its access permissions, improving the security of data in memory, thereby reducing the security risks of the intelligent agent and improving its reliability.

[0058] Based on the same inventive concept, embodiments of the present invention also provide a computer program product, including a computer program, which, when executed, implements the memory permission control method for an intelligent agent as described in any of the foregoing embodiments.

[0059] Based on the same inventive concept, embodiments of the present invention also provide an electronic device, including a memory and a processor, wherein the memory stores a program, and the processor calls the program stored in the memory to execute the memory access control method for an intelligent agent as described in any of the foregoing embodiments.

[0060] Based on the same inventive concept, embodiments of the present invention also provide a storage medium storing a program, which, when executed, implements the memory permission control method for an intelligent agent as described in any of the foregoing embodiments.

[0061] The foregoing describes multiple embodiments of the present invention. The optional methods described in each embodiment can be combined and cross-referenced without conflict, thereby extending to a variety of possible embodiments. These can all be considered as embodiments disclosed or made public by the present invention.

[0062] While the embodiments of the present invention have been disclosed above, the present invention is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the present invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A method for controlling memory access permissions for an intelligent agent, characterized in that, Processors used in intelligent agents include: When the activation of a third-party skill module of the intelligent agent is detected, a pre-configured permission configuration table is loaded; the permission configuration table stores the access permissions of the third-party skill module to memory regions with different security levels; During the operation of the third-party skill module, based on the permission configuration table and the access behavior of the third-party skill module to the memory area, it is determined whether there is any access behavior that exceeds its access permissions; if so, access is denied.

2. The memory access control method for an intelligent agent as described in claim 1, characterized in that, The intelligent agent also includes: a core program module; the permission configuration table also stores the access permissions of the core program module to memory regions with different security levels; Before loading the pre-configured permission configuration table when the activation of the third-party skill module of the intelligent agent is detected, the process also includes: forming the permission configuration table; The steps for forming the permission configuration table include: The security level of each memory region is determined based on the stored data objects in each memory region; Based on the security level of the memory region, determine the access permissions of the core program module and the third-party skill module under different access behaviors to the memory region; The permission configuration table is formed by storing memory regions with different security levels and their corresponding stored data objects, as well as the access permissions of core program modules and third-party skill modules to memory regions with different security levels.

3. The memory access control method for an intelligent agent as described in claim 2, characterized in that, Also includes: When the core program module of the intelligent agent is detected to be started, a pre-configured permission configuration table is loaded; During the operation of the core program module, based on the permission configuration table and the core program module's access behavior to the memory region, it is determined whether there is any access behavior that exceeds its access permissions. If it exists, access is denied.

4. The memory access control method for an intelligent agent as described in claim 2, characterized in that, The stored data objects include: local key sensitive data, external skill data, and user general data; The step of determining the security level of each memory region based on the stored data objects in each memory region includes: When the memory area stores local critical and sensitive data, the security level of the memory area is determined to be the first security level, and the memory area is designated as the core sensitive area; when the memory area stores ordinary user data, the security level of the memory area is determined to be the second security level, and the memory area is designated as the ordinary user area; when the memory area stores external skill data, the security level of the memory area is determined to be the third security level, and the memory area is designated as the skill operation area. In terms of security level, the first security level is higher than the second security level, and the second security level is higher than the third security level.

5. The memory access control method for an intelligent agent as described in claim 3, characterized in that, The access behaviors include: reading, writing, and deleting; The step of determining the access permissions of the core program module and the third-party skill module under different access behaviors of the memory region based on the security level of the memory region includes: When the security level of the memory region is the first security level, the core program module is determined to have access to the memory region, while the third-party skill module is determined to have access to the memory region but not. When the security level of the memory region is the second security level, the access permission of the core program module to the memory region is determined to be accessible, and the access permission of the third-party skill module to the memory region is readable, but not writeable or deletable; When the security level of the memory region is the third security level, the access rights of the core program module to the memory region are determined to be accessible, and the access rights of the third-party skill module to the memory region are also determined to be accessible.

6. The memory access control method for an intelligent agent as described in claim 5, characterized in that, During the operation of the third-party skill module, based on the permission configuration table and the third-party skill module's access behavior to the memory region, it is determined whether there is any access behavior exceeding its access permissions, including: When the third-party skill module accesses the core sensitive area during operation, it is determined that the third-party skill module has access behavior exceeding its access permissions during operation. When the third-party skill module accesses the ordinary user area during operation, it is determined whether the access behavior is the third-party skill module reading data in the ordinary user area. If yes, it is determined that the third-party skill module does not have access behavior exceeding its access permissions during operation; otherwise, it is determined that the third-party skill module has access behavior exceeding its access permissions during operation. When the third-party skill module accesses the skill operation area during operation, it is determined that the third-party skill module does not have any access behavior exceeding its access permissions during operation.

7. The memory access control method for an intelligent agent as described in claim 5, characterized in that, During the operation of the core program module, based on the permission configuration table and the core program module's access behavior to the memory region, it is determined whether there is any access behavior exceeding its access permissions, including: When the core program module accesses the core sensitive area, the ordinary user area, and the skill operation area during operation, there is no access behavior that exceeds its access permissions.

8. The memory access control method for an intelligent agent as described in any one of claims 1-7, characterized in that, When access is denied, the event corresponding to the denied access is also stored in the security log.

9. A memory access control device for an intelligent agent, characterized in that, Processors used in intelligent agents include: A loading module is used to load a pre-configured permission configuration table when the activation of a third-party skill module of the intelligent agent is detected; the permission configuration table stores the access permissions of the third-party skill module to memory regions of different security levels; The access control module is used to determine, during the operation of the third-party skill module, whether there is any access behavior that exceeds its access permissions based on the access configuration table and the access behavior of the third-party skill module to the memory area; if so, the access is denied.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed, it implements the memory access control method for an intelligent agent as described in any one of claims 1-8.

11. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a program, and the processor calls the program stored in the memory to execute the memory access control method for an intelligent agent as described in any one of claims 1-8.

12. A storage medium, characterized in that, The storage medium stores a program, which, when executed, implements the memory access control method for an intelligent agent as described in any one of claims 1-8.

Images