Method, device, medium and program product for implementing multi-level cloud resource management

Abstract

The application provides a method, device, medium and program product for implementing multi-level cloud resource management, which comprises the following steps: receiving a request of a senior user for requesting a junior user to execute an action through a user request agent; obtaining a trust score of the senior user at the current time through a dynamic trust engine, and determining a policy related to the action; modeling real-time credibility of the senior user when proxying the junior user resources as a Bayesian learning process by the dynamic trust engine; the trust score is subject to Beta distribution; according to a pre-constructed objective function and constraint condition, calculating an activation vector of an optimal policy related to the operation to obtain an optimal solution; when the optimal solution is greater than or equal to a predetermined utility threshold, determining whether the permission range of the senior user contains the operation according to the activation vector, and if yes, calculating a token survival time of the operation b and issuing a corresponding dynamic proxy token. By using the above technical scheme, fine-grained and dynamic multi-level cloud resource management can be realized for the proxy behavior.

Description

Technical Field

[0001] This invention relates to the field of cloud computing security and distributed resource management technology, and in particular to a method, apparatus, medium and program product for implementing multi-level cloud resource management. Background Technology

[0002] With the widespread adoption of cloud computing services, large cloud service providers (CSPs) commonly employ multi-tier distribution models to expand their market share. In this model, upstream users (hosting users) can allocate and license their cloud resources (such as compute instances, storage buckets, virtual networks, etc.) to downstream users (resellers). Downstream resellers can independently manage their allocated resources, such as creating and managing their customers' private IP (PBX) systems.

[0003] In actual operation, two complex collaboration needs have emerged: Operational assistance requirement: Lower-level resellers may lack sufficient technical capabilities and require intervention from higher-level hosting users to assist with resource management and operation. This requires lower-level resellers to temporarily grant higher-level users specific operation permissions for their resources (such as "login to the PBX management interface" and "create a new PBX instance").

[0004] Outsourced operations needs: In order to be more deeply involved in the business, the upper-level Hosting User may want to create and manage resources directly from the perspective and resource quota of the lower-level Reseller, for example, to help the lower-level Reseller quickly initialize a customer environment.

[0005] Existing technical solutions typically address this issue in the following ways: Credential sharing: Lower-level resellers directly provide their account passwords or API keys to their superiors. This method has extremely low security, violates the principle of least privilege, and cannot perform fine-grained access control and operation auditing. Once a security issue occurs, responsibility cannot be traced.

[0006] Static role-playing: This method utilizes mechanisms such as AWS IAM Roles, where a subordinate creates a role and grants permissions to the superior to assume that role. However, permissions in this approach are statically predefined and lack flexibility. When the superior needs to perform a new operation that was not pre-defined in the role, the subordinate must modify the role policy, resulting in a cumbersome process that cannot meet the needs of dynamic and ever-changing operational scenarios.

[0007] A simple policy switch: The system provides a switch to "enable / disable" global passwordless login (SSO). When enabled, the superior gains access to all resources of the subordinate; when disabled, the subordinate has no access at all. This "all or nothing" binary model is too coarse, either leading to excessive permissions and security risks, or failing to meet temporary, fine-grained collaboration needs.

[0008] The shortcomings of existing technologies can be summarized as follows: The conflict between granularity and flexibility in permissions: It is impossible to provide flexible, on-demand, temporary advanced permissions while ensuring security (least privileges); static roles cannot cover all temporary operation and maintenance scenarios, while credential sharing and global on / off permissions are too broad. Lack of dynamic trust mechanism: The granting of permissions is static and cannot be dynamically adjusted based on the historical proxy behavior of the superior user (such as the success rate of operation and whether security rules are triggered); a trusted superior and a superior who is acting as a proxy for the first time have the same static permissions, which is unreasonable and risky. Insufficient intelligence in policy conflict and adjudication: When the lower level sets multiple policies for users (such as one that is specific to a certain operation and another that is a global policy), the existing system usually adopts a simple "rejection priority" or "sequential matching" logic, which cannot make the optimal decision that comprehensively considers trust, risk and utility under multiple constraints. Weak security and auditability: Credential sharing leads to account proliferation; static roles make it difficult to track which superior performs a specific operation; the system lacks a complete and non-repudiable audit capability for the proxy operation chain.

[0009] Therefore, there is an urgent need in this field for a multi-level cloud resource management solution to address one or more of the aforementioned problems. Summary of the Invention

[0010] The embodiments of the present invention provide a method, apparatus, medium and program product for implementing multi-level cloud resource management. By continuously evaluating the trust level of superior users through statistical methods, the risk of proxy behavior is continuously assessed, and fine-grained and dynamic multi-level cloud resource management of proxy behavior is achieved.

[0011] It reduces the power consumption of the embedded subsystem while performing preset functions in real time.

[0012] To achieve the above objectives, on the one hand, a method for implementing multi-level cloud resource management is provided, including: Receive a request from the superior user u_h to proxy the subordinate user u_r to execute the first action b; The dynamic trust engine obtains the trust score T(u_h, u_r) of the superior user u_h at the current moment, and determines all strategies related to the first action based on the pre-set correspondence between actions and strategies. The dynamic trust engine models the real-time credibility of the superior user when acting as an agent for the resources of the subordinate user as a Bayesian learning process. The real-time credibility is indicated by a trust score, which follows a Beta distribution. The parameter α of the Beta distribution is the cumulative measure of historical successful operations, and the parameter β is the cumulative measure of historical failed or risky operations. The trust score at the current moment is the expected value of its posterior Beta distribution. Based on the pre-constructed objective function U(x) and predetermined constraints, calculate the activation vector x* of the optimal policy associated with the first action b, and obtain the optimal solution U(x*), where the optimal policy is the policy that maximizes the value of the objective function U(x); The static weights are preset for strategy i; Let be the dynamic trust score associated with policy i; where policy i is a specific policy of the lower-level user u_r against the upper-level user u_h. T(u_h, u_r); N is the total number of all strategies related to the first action b; To optimize variables; This is a risk penalty item; This is a preset global risk aversion coefficient used to indicate the degree of risk aversion, with a value greater than 0.

[0013]

[0014] This is a pre-defined inherent risk score for the first action b; When the optimal solution U(x*) is greater than or equal to the predetermined utility threshold, the permission scope of the superior user u_h is determined according to the activation vector x*. If the first action b is included, the token lifetime of the first action b is calculated and the corresponding dynamic proxy token is issued. The token time-to-live (TTL) for the first action b is calculated using the following formula:

[0015] in, This sets the maximum allowed token lifespan. This is the decay rate coefficient, and its value is greater than 0.

[0016] Preferably, in the method described, when policy i is a global policy, This is the average trust score of all superior users who have acted as agents for the aforementioned subordinate users.

[0017] Preferably, the method further includes: The resource manager verifies the dynamic proxy token and executes the first action b, generating an audit log.

[0018] Preferably, the method further includes: Multidimensional features are extracted from audit logs in real time to form a feature vector. The multidimensional features include: short-term operation frequency, deviation of operation time from the predetermined user's normal working time, resource consumption rate, and entropy of operation sequence. The anomaly score A for this operation is calculated using the isolated forest algorithm on the feature vector. When the anomaly score A is greater than the predetermined anomaly operation threshold, the operation is determined to be an anomaly or risky operation, and the result of the operation is determined to be a failure when trust is updated; otherwise, the operation is determined to be a successful operation, and the result of the operation is determined to be a success when trust is updated.

[0019] Preferably, the method further includes: Based on the result of this operation, the dynamic trust engine uses the Bayesian update formula to refresh the (α, β) parameters of the Beta distribution, thus completing a learning loop.

[0020] Preferably, in the method, the resource consumption rate includes the rate of newly created PBXs.

[0021] Preferably, in the method, the preset constraints include one or more of the following: Logical mutual exclusion constraints are used to constrain logically conflicting strategies. The request action matching constraint is used to constrain the number of policies that must be activated to allow an action. The minimum trust threshold constraint is used to constrain the dynamic trust score of a policy whose action can be activated to be greater than or equal to the minimum trust threshold.

[0022] On the other hand, an apparatus for implementing multi-level cloud resource management is also provided, including a memory and a processor, wherein the memory stores at least one program, which is executed by the processor to implement the steps of the method for implementing multi-level cloud resource management as described above.

[0023] In another aspect, a computer-readable storage medium is also provided, wherein at least one program is stored therein, the at least one program being executed by a processor to implement the steps of the method for implementing multi-level cloud resource management as described above.

[0024] In another aspect, a computer program product is also provided, comprising a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the method for implementing multi-level cloud resource management as described above.

[0025] The above technical solution has the following technical effects: The technical solution of this invention uses a Bayesian-updated dynamic trust model to make the granting of agent behavior permissions no longer static, but adaptively adjusted according to the history of user behavior, thus realizing intelligent risk control from static to dynamic adaptation.

[0026] In a further embodiment, by modeling policy conflict resolution as a constrained optimization problem and introducing risk and utility quantification, the system is able to make globally optimal and refined permission decisions, rather than a simple "yes / no" judgment, thus realizing cloud resource management from binary decision-making to optimal decision-making.

[0027] In a further embodiment, by incorporating trust scores and risk levels into the dynamic calculation of TTL, security policies are closely linked to real-time risks, and stricter restrictions, such as shorter token validity periods, are automatically imposed on high-risk operations, thus realizing a shift from fixed policies to risk-aware policies.

[0028] The technical solution of this invention provides a multi-level cloud resource management method based on random trust assessment and optimized role agent. It creatively integrates Bayesian statistics, integer programming, anomaly detection machine learning algorithms with the specific application scenario of cloud resource access control, forming a scientific, rigorous, and computable automated security management system for cloud resources. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating a method for implementing multi-level cloud resource management according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating a method for implementing multi-level cloud resource management according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of a device for implementing multi-level cloud resource management according to an embodiment of the present invention. Detailed Implementation

[0030] To further illustrate the various embodiments, the present invention provides accompanying drawings. These drawings are part of the disclosure of the present invention, primarily used to illustrate the embodiments and to explain the operating principles of the embodiments in conjunction with the relevant descriptions in the specification. With reference to these drawings, those skilled in the art should be able to understand other possible implementations and the advantages of the present invention. Components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0031] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0032] Example 1: Figure 1 This is a flowchart illustrating a method for implementing multi-level cloud resource management according to an embodiment of the present invention. Figure 1 The method for implementing multi-level cloud resource management in this embodiment includes: S1, Receive the request from the superior user u_h to proxy the subordinate user u_r to perform the first action b; S2, the current trust score T(u_h, u_r) of the superior user u_h is obtained through the dynamic trust engine, and all strategies related to the first action b are determined according to the pre-set correspondence between actions and strategies; the dynamic trust engine models the real-time credibility of the superior user when acting as an agent for the resources of the subordinate user as a Bayesian learning process; the trust score follows a Beta distribution, where the parameter α of the Beta distribution is the cumulative measure of historical successful operations, and β is the cumulative measure of historical failed or risky operations; the trust score at the current moment is taken as the expected value of its posterior Beta distribution; S3. Based on the pre-constructed objective function U(x) and predetermined constraints, calculate the activation vector x* of the optimal policy associated with the first action b, and obtain the optimal solution U(x*), where the optimal policy is the policy that maximizes the value of the objective function U(x). The static weights are preset for strategy i; Let be the dynamic trust score associated with policy i; where policy i is a specific policy of the lower-level user u_r against the upper-level user u_h. T(u_h, u_r); N is the total number of strategies related to the first action b; To optimize variables; This is a risk penalty item; This is a preset global risk aversion coefficient used to indicate the degree of risk aversion, with a value greater than 0.

[0033]

[0034] The inherent risk score is preset for the first action b; S4. When the optimal solution U(x*) is greater than or equal to the predetermined utility threshold, determine whether the permission scope of the superior user u_h includes the first action b based on the activation vector x*. If it includes the first action, calculate the token lifetime of the first action b and issue the corresponding dynamic proxy token. The token time-to-live (TTL) for the first action b is calculated using the following formula:

[0035] in, This sets the maximum allowed token lifespan. This is the decay rate coefficient, and its value is greater than 0.

[0036] In one specific implementation, when policy i is a global policy, This is a statistical value of the trust score of all superior users who have acted as agents for the aforementioned subordinate users; this statistical value may be, for example, the average of the trust scores; or, other statistical values ​​may be used depending on the needs and scenarios.

[0037] Preferably, the method in this embodiment of the invention further includes: the resource manager verifying the dynamic proxy token and performing the first action b to generate an audit log.

[0038] Preferably, the method in this embodiment of the invention further includes: Multidimensional features are extracted from audit logs in real time to form a feature vector. These multidimensional features include: short-term operation frequency, deviation of operation time from the scheduled user working time, resource consumption rate, and entropy of operation sequence. In one specific implementation, the resource consumption rate includes: the rate of new PBX creation; the rate of new PBX creation = number of new PBXs created / time. The isolation forest algorithm is used to calculate the anomaly score A for this operation based on the feature vector; When the anomaly score A is greater than the predetermined anomaly operation threshold, the operation is determined to be an anomaly or risky operation, and the result of the operation is determined to be a failure when trust is updated; otherwise, the operation is determined to be a successful operation, and the result of the operation is determined to be a success when trust is updated.

[0039] Preferably, the method in the embodiment of the present invention further includes: the dynamic trust engine uses the Bayesian update formula to refresh the (α, β) parameters of the Beta distribution based on the result of the current operation, thus completing a learning loop.

[0040] Preferably, the aforementioned preset constraints include one or more of the following: Logical mutual exclusion constraints are used to constrain logically conflicting strategies. The request action matching constraint is used to constrain the number of policies that must be activated to allow an action. The minimum trust threshold constraint is used to constrain the dynamic trust score of a policy whose action can be activated to be greater than or equal to the minimum trust threshold.

[0041] Example 2: The core of this embodiment of the invention lies in the introduction of a dynamic trust engine, which continuously evaluates the risk of proxy behavior through statistical methods and uses optimization algorithms to calculate the optimal permission allocation strategy under multiple constraints.

[0042] Figure 2 This is a flowchart illustrating the method for implementing multi-level cloud resource management according to an embodiment of the present invention. Figure 2 This illustrates the interaction and closed-loop feedback between the dynamic trust engine and other program modules that implement corresponding functions. The functions and implementations of each relevant module of the method in this embodiment, as well as the overall workflow, are described below.

[0043] 1. Dynamic Trust Assessment Model The dynamic trust engine, or dynamic trust evaluation model, is used to quantify the real-time trustworthiness of a higher-level user u_h when acting as an agent for lower-level user u_r's resources. In this embodiment of the invention, real-time trustworthiness is modeled as a Bayesian learning process.

[0044] Trust score T(u_h, u_r) is used to identify the above real-time trustworthiness: Definition: A continuous random variable taking values ​​in the interval [0, 1]. T = 1 represents complete trust, and T = 0 represents complete untrust. In one specific implementation, the initial value T_0 of the trust score is set to 0.5, representing neutral trust.

[0045] Probabilistic model: Assuming that the trust score and trust level T follow a Beta distribution, i.e., T Beta(α, β). The Beta distribution is chosen because it is the conjugate prior of the binomial and Bernoulli distributions, making it well-suited for probabilistic modeling of binary data such as success / failure.

[0046] Parameter explanation: α can be regarded as the cumulative evidence of historical successful operations; β can be regarded as the cumulative evidence of historical failed or risky operations. In a specific implementation, the initial values ​​of α and β are set to α_0 = 1 and β_0 = 1, corresponding to a uniform prior.

[0047] Bayesian update formula: Each agency operation and its subsequent evaluation is considered a Bernoulli trial. The trial result is determined by a risk behavior detection model. o = 1 (Success): The operation was completed without triggering any risk warnings.

[0048] o = 0 (failure): The operation failed, was actively interrupted, or a risk alarm was triggered. After observing the result o, the posterior parameters of the Beta distribution are updated as follows: text If o = 1 (success): α_{new} = α_{old} + 1 β_{new} = β_{old} If o = 0 (failure): α_{new} = α_{old} β_{new} = β_{old} + 1 That is: if the result is successful, the updated parameter α value is equal to the original parameter α value plus 1; the updated parameter β value is equal to the original parameter β value; if the result is unsuccessful, the updated parameter α value is equal to the original parameter α value; the updated parameter β value is equal to the original parameter β value plus 1.

[0049] Trust score calculation: The trust score T(u_h, u_r) at the current time is the expected value of its posterior Beta distribution: T(u_h, u_r) = E[Beta(α, β)] = α / (α + β) Trust uncertainty: The variance Var of the posterior distribution can measure the uncertainty of trust assessment. When uncertainty is high, the system can adopt a more conservative strategy, where: Var = αβ / [(α+β) 2 ×(α+β+1)] 2. Strategy Conflict Resolution and Optimal Authority Calculation When a lower-level user u_r sets multiple potentially conflicting ProxyPolicies, examples of policy conflicts include: one allowing PBX creation while another globally prohibits all write operations; or the simultaneous existence of global SSO and precise policies. In such cases, the system needs intelligent decision-making. This embodiment of the invention formalizes this problem as a constrained optimization problem.

[0050] Optimize variables definition: Suppose there are N strategies associated with the current request (u_h, u_r, b). Define a binary decision vector x = [x1, x2, ..., x...]. N ] T ∈ {0, 1} N . = 1 indicates that strategy i is activated in this request; otherwise, it is not activated.

[0051] Objective function (maximizing trust-utility):

[0052] The static weight of strategy i is preset by the lower-level user u_r when creating the strategy, reflecting the importance they attach to this strategy item. In one specific implementation, the strategy weight for core business is set to 1.0, and the weight for non-core strategies is set to 0.5.

[0053] : The dynamic trust score associated with policy i. For a policy targeting a specific superior, T_i = T(u_h,u_r); in one specific implementation, for a global policy, T_i can be the average trust score of all superiors who have acted as agents.

[0054] : Optimize variables.

[0055] Risk penalty items. An important design element is:

[0056] in, It is the inherent risk score of operation b. In a specific implementation, it is preset by the administrator. For example, read-only operation is 0.1 and delete operation is 1.0.

[0057] λ: Global risk aversion coefficient, λ>0. This parameter controls the overall system's aversion to risk. The larger λ is, the heavier the penalty for high-risk operations under low trust levels.

[0058] Constraints: Logical mutual exclusion constraint: For a logically conflicting strategy pair (i, j), x i + x j <= 1. For example, if one policy i is "Action: *", "Effect: Deny", and another policy j is "Action: CreatePBX", "Effect:Allow", and policy i is to deny and policy j is to allow, then they are mutually exclusive.

[0059] Request action matching constraint: Define a set V = {i | b ∈ Actions_i}, which is the policy index of all allowed actions b. Then the constraint is: This means that at least one policy that allows the action is activated; otherwise, the request is rejected directly due to lack of permission.

[0060] Minimum Trust Threshold Constraint: For certain extremely high-risk operations, the policy that activates it must satisfy the following condition. That is, greater than or equal to the predetermined trust score threshold. This can be expressed as .

[0061] Solution algorithm: The problem described above is a binary quadratic programming problem with linear constraints. For smaller problems, exact algorithms such as branch and bound can be used. For large-scale problems, metaheuristic algorithms such as simulated annealing or genetic algorithms can be used for efficient approximate solutions. The solver outputs the optimal solution x*, which determines the final set of policies to be activated.

[0062] 3. Enhancements to Dynamic Proxy Tokens (DPT) The scope field of the DPT token consists of all activated policies {i | The set of permissions is the union of the permissions of 1.

[0063] Dynamic calculation of token time-to-live (TTL): The token's lifespan is negatively correlated with trust score and operational risk level, employing an exponential decay model:

[0064] : The maximum token lifespan allowed by the system.

[0065] Attenuation rate coefficient >0 is used to control the TTL's sensitivity to trust and risk.

[0066] T(u_h, u_r): The current dynamic trust score.

[0067] The inherent risk score of operation b, whose value is preset.

[0068] Formula Interpretation: When the trust level T is very low or the operational risk is very high, the value of the exponent term will become smaller, resulting in a significant reduction in TTL, causing the token to expire faster and limiting potential damage.

[0069] 4. Risk Behavior Detection Model This model is implemented using machine learning algorithms such as the Isolation Forest algorithm, which serves as a data source for trust updates and is used to determine the success or failure of a single operation in real time.

[0070] Feature engineering: Extracting multi-dimensional features from the audit log stream in real time to construct a feature vector F: F1: Short-term operation frequency, for example, the number of operations in the past minute; F2: Deviation between operation time and user's normal working time.

[0071] F3: Resource consumption rate, for example, the number of newly created PBXs per time.

[0072] F4: Entropy of the operation sequence, used to identify differences from historically common operation sequences.

[0073] Anomaly detection: The isolation forest algorithm is adopted for this purpose; this algorithm is suitable for unsupervised anomaly detection of continuous data, and calculates the anomaly score A ∈ [0, 1] for each sample, that is, each operation, by constructing isolation trees.

[0074] Decision rule: Set the anomaly threshold A_threshold. If A > A_threshold for a certain operation, it is considered an abnormal / risky operation, and o = 0 is recorded during trust update, that is, the operation fails.

[0075] The overall workflow of the method in this embodiment of the present invention includes the following steps: Request: u_h requests to perform an action in the identity of u_r; Trust query and policy collection: The authentication gateway obtains T(u_h, u_r) from the dynamic trust engine, and queries all policies related to the operation b; in one implementation, it queries from a preset policy library. Optimization solution: The system constructs the objective function U(x) and constraint conditions, and calls the solver to calculate the optimal policy activation vector x*.

[0076] Conflict adjudication and token generation: If the optimization problem has no feasible solution or the optimal solution U(x*) < U_min, where U_min is a preset minimum utility threshold, the request is rejected.

[0077] If passed, the Security Token Service (STS) issues a DPT token according to the permission scope (scope) calculated based on x* and the expiration time calculated by the dynamic TTL formula. The permission scope indicates whether u_h is allowed to act as u_r for operation b.

[0078] Operation execution and auditing: The resource manager verifies the DPT token and executes the operation, generating a detailed audit log (AuditLog).

[0079] Feedback and update: The risk behavior detection model analyzes the AuditLog and calculates the anomaly score A corresponding to this operation.

[0080] Based on A and the operation result, determine whether the result of this trial is successful or recognized.

[0081] The dynamic trust engine uses the Bayesian update formula to refresh the (α, β) parameters of T(u_h, u_r) according to the value of o, completing a learning closed-loop.

[0082] Embodiment 3: The present invention also provides a device for implementing multi-level cloud resource management, such as Figure 3As shown, the device includes a processor 301, a memory 302, a bus 303, and a computer program stored in the memory 302 and executable on the processor 301. The processor 301 includes one or more processing cores. The memory 302 is connected to the processor 301 via the bus 303. The memory 302 is used to store program instructions. When the processor executes the computer program, it implements the steps in the above-described method embodiment of Embodiment 1 of the present invention.

[0083] Furthermore, as an executable solution, the device can be a computer unit, which may be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer unit may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above-described structure of the computer unit is merely an example and does not constitute a limitation on the computer unit. It may include more or fewer components, or combine certain components, or use different components. For example, the computer unit may also include input / output devices, network access devices, buses, etc., and this embodiment of the invention does not limit this.

[0084] Furthermore, as an executable solution, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the computer unit, connecting various parts of the entire computer unit via various interfaces and lines.

[0085] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer unit by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0086] Example 4: The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described in the embodiments of the present invention.

[0087] If the modules / units integrated in the computer unit are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0088] Example 5: The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps described above.

[0089] Although the invention has been specifically shown and described in conjunction with preferred embodiments, those skilled in the art should understand that various changes in form and detail may be made to the invention without departing from the spirit and scope of the invention as defined in the appended claims, all of which shall be within the scope of protection of the invention.

Claims

1. A method for implementing multi-level cloud resource management, characterized in that, include: Receive a request from the superior user u_h to proxy the subordinate user u_r to execute the first action b; The dynamic trust engine obtains the trust score T(u_h, u_r) of the superior user u_h at the current moment, and determines all strategies related to the first action based on the pre-set correspondence between actions and strategies. The dynamic trust engine models the real-time credibility of the superior user when acting as an agent for the resources of the subordinate user as a Bayesian learning process. The real-time credibility is indicated by a trust score, which follows a Beta distribution. The parameter α of the Beta distribution is the cumulative measure of historical successful operations, and the parameter β is the cumulative measure of historical failed or risky operations. The trust score at the current moment is the expected value of its posterior Beta distribution. Based on the pre-constructed objective function U(x) and predetermined constraints, calculate the activation vector x* of the optimal policy associated with the first action b, and obtain the optimal solution U(x*), where the optimal policy is the policy that maximizes the value of the objective function U(x); The static weights are preset for strategy i; Let be the dynamic trust score associated with policy i; where policy i is a specific policy of the lower-level user u_r against the upper-level user u_h. T(u_h, u_r); N is the total number of all strategies related to the first action b; To optimize variables; This is a risk penalty item; This is a preset global risk aversion coefficient used to indicate the degree of risk aversion, with a value greater than 0. This is a pre-defined inherent risk score for the first action b; When the optimal solution U(x*) is greater than or equal to the predetermined utility threshold, the permission scope of the superior user u_h is determined according to the activation vector x*. If the first action b is included, the token lifetime of the first action b is calculated and the corresponding dynamic proxy token is issued. The token time-to-live (TTL) for the first action b is calculated using the following formula: in, This sets the maximum allowed token lifespan. This is the decay rate coefficient, and its value is greater than 0.

2. The method according to claim 1, characterized in that, When policy i is a global policy This is the average trust score of all superior users who have acted as agents for the aforementioned subordinate users.

3. The method according to claim 1, characterized in that, Also includes: The resource manager verifies the dynamic proxy token and executes the first action b, generating an audit log.

4. The method according to claim 3, characterized in that, Also includes: Multidimensional features are extracted from audit logs in real time to form a feature vector. The multidimensional features include: short-term operation frequency, deviation of operation time from the predetermined user's normal working time, resource consumption rate, and entropy of operation sequence. The anomaly score A for this operation is calculated using the isolated forest algorithm on the feature vector. When the anomaly score A is greater than the predetermined anomaly operation threshold, the operation is determined to be an anomaly or risky operation, and the result of the operation is determined to be a failure when trust is updated; otherwise, the operation is determined to be a successful operation, and the result of the operation is determined to be a success when trust is updated.

5. The method according to claim 4, characterized in that, Also includes: Based on the result of this operation, the dynamic trust engine uses the Bayesian update formula to refresh the (α, β) parameters of the Beta distribution, thus completing a learning loop.

6. The method according to claim 4, characterized in that, The resource consumption rate includes the rate at which new PBXs are created.

7. The method according to claim 1, characterized in that, The preset constraints include one or more of the following: Logical mutual exclusion constraints are used to constrain logically conflicting strategies. The request action matching constraint is used to constrain the number of policies that must be activated to allow an action. The minimum trust threshold constraint is used to constrain the dynamic trust score of a policy whose action can be activated to be greater than or equal to the minimum trust threshold.

8. An apparatus for implementing multi-level cloud resource management, characterized in that, The system includes a memory and a processor, the memory storing at least one program that is executed by the processor to implement the steps of the method for implementing multi-level cloud resource management as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The storage medium stores at least one program, which is executed by a processor to implement the steps of the method for implementing multi-level cloud resource management as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method for implementing multi-level cloud resource management as described in any one of claims 1 to 7.

Images