Security access control method based on zero-trust model

Abstract

The invention discloses a security access control method based on a zero-trust model, which comprises the following steps: collecting multi-source context information corresponding to an access request, and generating a context information set; performing fuzzy coding processing on the context information set to obtain a fuzzy context variable set; constructing an improved fuzzy Bayesian network model, and forming a context adaptive inference network; inputting the fuzzy context variable set into a context adaptive reasoning network, executing a fuzzy reasoning operation, and outputting an access risk level assessment result; generating an access control instruction through an access control decision mapping module based on the access risk level evaluation result; executing access response processing according to the access control instruction, and generating an access processing result; and jointly constructing strategy feedback information by using an access risk level evaluation result and an access processing result, and updating the context adaptive inference network. According to the invention, the improved fuzzy Bayesian network is adopted, and multi-context adaptive access control is realized.

Description

Technical Field

[0001] This invention relates to the field of network security technology, and in particular to a secure access control method based on a zero-trust model. Background Technology

[0002] In traditional network security architectures, access control mechanisms are typically based on static trust boundary models, including Role-Based Access Control (RBAC), Attribute-Based Access Control (ABAC), and Access Control Lists (ACLs). While these methods offer some protection in closed network environments, their static authorization mechanisms have become increasingly inflexible, lacking granularity, and exhibiting slow response times in the context of multi-device access, widespread cloud computing deployments, and frequent changes in personnel roles. For example, RBAC relies on predefined role and permission mappings, lacking awareness of real-time behavior and contextual states, making it difficult to adjust in real-time when access behavior is abnormal or device status is untrusted. Although ABAC introduces attribute factors, its attribute matching logic is usually statically configured, unable to adapt to contextual changes, and struggles to handle non-linear dependencies between attributes.

[0003] In recent years, the zero-trust security model has been gradually proposed and deployed in practical systems. Its core concept is to no longer assume the internal network is trustworthy, but rather to treat all access requests as potential risks, requiring authentication, device verification, and environmental assessment before access. Although this model strengthens the necessity of pre-access verification in its concept, many challenges remain in practical implementation. Current zero-trust implementations mostly employ fixed rule bases, access log pattern recognition, or single-factor scoring mechanisms, which have limited ability to model multi-source contextual information and struggle to comprehensively consider contextual factors such as user identity, device status, network environment, operational behavior, and spatiotemporal characteristics. Furthermore, traditional methods struggle to establish effective modeling structures for complex coupling relationships between contextual factors, such as the potential causal relationship between device status and operational behavior. In addition, current solutions lack the ability to model historical access behavior with time sensitivity, making it difficult to dynamically perceive changes in access trustworthiness based on the temporal evolution of access behavior.

[0004] Existing research has attempted to introduce artificial intelligence algorithms to assist access control, such as fuzzy logic, Bayesian networks, or neural networks, but shortcomings remain: fuzzy logic rules are difficult to update dynamically and lack closed-loop linkage with environmental feedback; traditional Bayesian networks have fixed structures and cannot dynamically adjust node relationships based on contextual states; risk assessment and control instructions are often separated, lacking a mechanism to feed execution results back to the evaluation model, thus limiting policy evolution capabilities. Furthermore, there is currently a lack of a method to input fuzzy contextual information into a structurally variable inference network and update conditional probability parameters and membership functions in real time based on access processing results.

[0005] Therefore, how to provide a secure access control method based on a zero-trust model is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a secure access control method based on a zero-trust model. This invention introduces an improved fuzzy Bayesian network model, which integrates fuzzy membership function modeling, time-sensitive fuzzy node structure and dynamic structure adjustment mechanism. It describes in detail how to dynamically assess access risks and output access control policies based on multi-source context information, and has the advantages of strong reasoning ability, fine context awareness and high policy adaptive evolution ability.

[0007] A secure access control method based on a zero-trust model according to an embodiment of the present invention includes the following steps:

[0008] Collect multi-source context information corresponding to access requests and generate a context information set;

[0009] Based on a preset set of fuzzy membership functions, the set of context information is subjected to fuzzy encoding to obtain a set of fuzzy context variables.

[0010] Construct an improved fuzzy Bayesian network model to form a context-adaptive inference network;

[0011] The set of fuzzy context variables is input into the context adaptive inference network, fuzzy inference operations are performed, and the access risk level assessment result is output.

[0012] Based on the access risk level assessment results, access control instructions are generated through the access control decision mapping module.

[0013] Execute access response processing according to access control instructions and generate access processing results;

[0014] The access risk level assessment results and access processing results are used together to construct policy feedback information and update the context-adaptive inference network.

[0015] Optionally, the context information set includes user identity attribute information, device attribute information, network environment attribute information, operation behavior characteristic information, and time and location attribute information.

[0016] Optionally, the step of performing fuzzy encoding on the context information set to obtain a fuzzy context variable set specifically includes:

[0017] Based on a preset set of fuzzy membership functions, fuzzification processing is performed on various attribute information in the context information set. The set of fuzzy membership functions includes identity credibility function, device credibility function, network risk function, and behavior anomaly function.

[0018] Input the user's identity attribute information into the identity credibility function to calculate the identity credibility level and the corresponding fuzzy membership degree;

[0019] Input the device attribute information into the device credibility function to calculate the device credibility level and the corresponding fuzzy membership degree;

[0020] Input the network environment attribute information into the network risk function to calculate the network risk level and the corresponding fuzzy membership degree;

[0021] Input the operational behavior feature information into the behavior anomaly function to calculate the behavior anomaly level and the corresponding fuzzy membership degree;

[0022] The identity trust level, device trust level, network risk level, and abnormal behavior level are organized into a set of fuzzy context variables.

[0023] Optionally, the construction of the improved fuzzy Bayesian network model to form a context-adaptive inference network specifically includes:

[0024] An improved fuzzy Bayesian network model is constructed, which includes a node set, a structural connection set, and a fuzzy conditional probability table set.

[0025] The node set takes the set of fuzzy context variables as input, and each node represents a fuzzy context variable.

[0026] Based on a preset set of causal dependencies, a structural connection set is constructed, which consists of multiple directed edges in the node set, forming a directed acyclic graph structure.

[0027] A time-sensitive fuzzy node structure is introduced, wherein the time-sensitive fuzzy node performs time-weighted processing on the historical state sequence and introduces a time decay factor on the fuzzy membership value of the node at the current time.

[0028] A context structure adaptation mechanism is constructed to dynamically adjust the set of structural connections based on the context state carried by the current access request, activating or blocking some edge connections to form an inference network structure.

[0029] Configure a fuzzy conditional probability table for a node, wherein the fuzzy conditional probability table defines the fuzzy output probability distribution of a child node under different combinations of parent node states;

[0030] The set of nodes, the set of structural connections, and the set of fuzzy conditional probability tables are integrated to construct a context-adaptive inference network.

[0031] Optionally, the introduction of the time-sensitive fuzzy node structure specifically includes:

[0032] A time-sensitive mechanism is introduced into each fuzzy node in the improved fuzzy Bayesian network model to construct a historical state sequence, which consists of multiple context observation times.

[0033] For each fuzzy context variable in the fuzzy context variable set at each time step, the fuzzy membership value is calculated to obtain the fuzzy state time series;

[0034] A time-weighting mechanism is introduced for fuzzy state time series, and each historical moment t is weighted according to its time position. i Assign weight values, which are calculated using an exponential decay method;

[0035] The fuzzy states of each fuzzy context variable are aggregated based on the time-weighted function to obtain the time-weighted fuzzy membership values:

[0036]

[0037] in, For fuzzy context variable v j The time-weighted fuzzy membership value at the current moment, k represents the historical state sequence. The time window length, where i represents the historical state sequence. The time index variable, t i For the i-th historical time point, w(t) i (t) represents time point t i The corresponding weight value, Fuzzy context variable v j At time t i The fuzzy membership value.

[0038] Optionally, the construction context structure adaptation mechanism specifically includes:

[0039] Receive the context state vector carried by the current access request, wherein the context state vector is composed of the time-weighted fuzzy membership values ​​of each fuzzy context variable in the fuzzy context variable set;

[0040] The directed edges in the structure connection set are matched and determined based on a preset set of structure activation rules.

[0041] For each directed edge, determine whether to activate the directed edge based on the matching results in the set of structural activation rules. If the current context state satisfies a certain activation rule, mark it as active; otherwise, mark it as passive.

[0042] A dynamic structural connection set is constructed, which consists of edge connections in the original structural connection set that satisfy the activation condition, and serves as the inference network structure in the current context state.

[0043] Optionally, the step of performing fuzzy inference and outputting the access risk level assessment result specifically includes:

[0044] The fuzzy context variable set is used as the input node variable and fed into the context adaptive inference network.

[0045] Based on the currently active dynamic structure connection set, perform graph traversal reasoning, combine the fuzzy conditional probability table of each node, and update the fuzzy membership degree of the child node state according to the combination of the parent node state membership degree;

[0046] A fuzzy level label set is generated at the inference terminal node. The fuzzy level label set includes low risk level, medium risk level and high risk level. The fuzzy membership degree value of each fuzzy level label is calculated to form a membership degree distribution set. The membership degree distribution set includes the membership degree value of low risk level, the membership value of medium risk level and the membership value of high risk level.

[0047] The fuzzy level label set and the membership degree distribution set are combined to form the access risk level assessment result.

[0048] Optionally, the generation of access control instructions through the access control decision mapping module specifically includes:

[0049] Receive access risk level assessment results, which consist of a fuzzy level label set L and a corresponding membership degree distribution set μ. L constitute;

[0050] Based on the membership distribution set, an access risk level judgment operation is performed. When the membership value of a high-risk level is greater than the membership values ​​of medium-risk and low-risk levels, the access request is determined to be of high risk level, and an "access denied" operation instruction is generated.

[0051] When the membership value of the medium-risk level is greater than that of the high-risk and low-risk levels, the access request is determined to be of medium-risk level, and an operation instruction of "restrict access scope" or "perform secondary verification" is generated.

[0052] When the membership value of a low-risk level is greater than that of a high-risk or medium-risk level, the access request is determined to belong to the low-risk level, and an "access allowed" operation instruction is generated.

[0053] The control actions corresponding to the access risk level judgment are used as the access control decision results, and then encapsulated to generate access control instructions.

[0054] Optionally, the access processing result includes control action result, trigger behavior record and context snapshot information; the control action result represents the response status after the access control command is executed; the trigger behavior record includes records of key behavioral events triggered during the access process and their execution status; the context snapshot information saves structured data of user identity attributes, device attributes, network environment attributes, operation behavior characteristics and time and location attributes corresponding to the execution of the access control action.

[0055] Optionally, the context-adaptive inference network specifically includes:

[0056] Construct a set of policy feedback information, where each policy feedback information is composed of the corresponding access risk level assessment result and access processing result;

[0057] Based on the policy feedback information set, update and improve the fuzzy conditional probability table set in the fuzzy Bayesian network model;

[0058] Based on the policy feedback information set, the fuzzy membership function set is adjusted, and the fuzzy level boundary value is offset and corrected according to the actual feedback during the update process.

[0059] By introducing context snapshot information from the strategy feedback information, a structural dynamic adjustment index function is constructed. The structural dynamic adjustment index function enables or disables some directed connection edges in the inference network based on the state combination of the current fuzzy context variable set, and generates a new structural connection set.

[0060] An updated context-adaptive inference network is constructed based on the updated fuzzy conditional probability table, fuzzy membership function set, and structural connection set.

[0061] The beneficial effects of this invention are:

[0062] This invention achieves standardized modeling of user identity attributes, device attributes, network environment, operational behavior characteristics, and time and location attributes by constructing a multi-source context information acquisition and fuzzy coding mechanism. By fuzzifying various attributes through a preset set of fuzzy membership functions, fuzzy context variables such as identity trust level, device trust level, network risk level, and abnormal behavior level are generated. This effectively improves the access control system's ability to express the access environment status and provides accurate input for subsequent risk assessment.

[0063] In the risk assessment process, this invention introduces an improved fuzzy Bayesian network model, combining a time-sensitive fuzzy node structure with a dynamic structural adjustment mechanism to construct a context-adaptive inference network. This inference network can not only perceive the temporal evolution trend of historical context states, but also activate or block some causal connection paths based on the current context state, thereby achieving dynamic inference of access risk levels. Compared with traditional models with fixed structures, this model has significant advantages in structural flexibility and inference robustness, enabling more accurate assessment of the risk level of access requests.

[0064] Furthermore, this invention designs a policy feedback mechanism based on access processing results. It jointly constructs feedback information from access risk assessment results and access response results, which is used to dynamically update the fuzzy conditional probability table and fuzzy membership function. The connection structure of the inference network is optimized through a structural dynamic adjustment function, thereby achieving continuous evolution of the access control policy. Through this closed-loop optimization process, the system can adapt to changing access environments during long-term operation, enhancing the adaptive and policy evolution capabilities of the access control system under a zero-trust architecture. Attached Figure Description

[0065] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0066] Figure 1 This is a flowchart of a secure access control method based on a zero-trust model proposed in this invention;

[0067] Figure 2 This is a schematic diagram of the structure of the improved fuzzy Bayesian network model in the secure access control method based on the zero-trust model proposed in this invention.

[0068] Figure 3 This is a schematic diagram of the time-weighted modeling process of time-sensitive fuzzy nodes in a secure access control method based on a zero-trust model proposed in this invention.

[0069] Figure 4 This is a schematic diagram of the dynamic structural connection adjustment of the context structure adaptive mechanism in the security access control method based on the zero trust model proposed in this invention. Detailed Implementation

[0070] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0071] refer to Figure 1-4 A secure access control method based on a zero-trust model includes the following steps:

[0072] Collect multi-source context information corresponding to access requests and generate a context information set;

[0073] Based on a preset set of fuzzy membership functions, the set of context information is subjected to fuzzy encoding to obtain a set of fuzzy context variables.

[0074] Construct an improved fuzzy Bayesian network model to form a context-adaptive inference network;

[0075] The set of fuzzy context variables is input into the context adaptive inference network, fuzzy inference operations are performed, and the access risk level assessment result is output.

[0076] Based on the access risk level assessment results, access control instructions are generated through the access control decision mapping module.

[0077] Execute access response processing according to access control instructions and generate access processing results;

[0078] The access risk level assessment results and access processing results are used together to construct policy feedback information and update the context-adaptive inference network.

[0079] This invention constructs a fuzzy context variable set based on multi-source context information and achieves dynamic reasoning and refined assessment of access risks through an improved fuzzy Bayesian network model. The system can perceive the comprehensive state of user identity, device status, network environment, and operational behavior, forming a structurally variable inference network that supports real-time decision-making in different contextual scenarios. Through a closed-loop feedback mechanism between access risk level and processing results, the system possesses adaptive optimization capabilities, continuously improving the accuracy and response efficiency of access control policies. This invention improves the precision, flexibility, and robustness of secure access control, making it suitable for access risk prevention and control in complex and dynamic network environments.

[0080] In this embodiment, the context information set includes user identity attribute information, device attribute information, network environment attribute information, operation behavior characteristic information, and time and location attribute information;

[0081] The user identity attribute information includes user identifier, authentication method, permission role, and historical access behavior characteristics; the device attribute information includes device type, operating system information, trusted startup status, terminal protection status, and access anomaly records; the network environment attribute information includes network access method, encryption status, network address, address reputation score, and historical network anomaly behavior; the operation behavior characteristic information includes target resource identifier, operation type, operation frequency, behavior path sequence, and operation time distribution; the time and location attribute information includes access timestamp, geographical location, time zone information, and location trust level.

[0082] This invention collects five types of contextual information—user identity, device attributes, network environment, operational behavior, and time and location—to construct a structured contextual information set. Multi-dimensional features are extracted and fuzzed, enhancing the comprehensiveness and precision of access risk perception. The various attribute information covers user authentication, device security status, network access security, operational behavior patterns, and geographic spatiotemporal consistency, aiding in the identification of potential risky access behaviors. Through multi-source information fusion and dynamic modeling mechanisms, this invention enhances the access control system's ability to discriminate abnormal behaviors in complex access scenarios, improving the accuracy of access control policies and the overall system security level.

[0083] In this embodiment, the step of performing fuzzy encoding on the context information set to obtain a fuzzy context variable set specifically includes:

[0084] Based on a preset set of fuzzy membership functions, fuzzification processing is performed on various attribute information in the context information set. The set of fuzzy membership functions includes identity credibility function, device credibility function, network risk function, and behavior anomaly function.

[0085] The user's identity attribute information is input into the identity credibility function to calculate the identity credibility level and the corresponding fuzzy membership degree. The identity credibility level includes high credibility, medium credibility and low credibility.

[0086] The device attribute information is input into the device credibility function to calculate the device credibility level and the corresponding fuzzy membership degree. The device credibility level includes credible, partially credible and untrustworthy.

[0087] The network environment attribute information is input into the network risk function to calculate the network risk level and the corresponding fuzzy membership degree. The network risk level includes low risk, medium risk and high risk.

[0088] The operational behavior feature information is input into the behavior anomaly function to calculate the behavior anomaly level and the corresponding fuzzy membership degree. The behavior anomaly level includes normal, deviation and anomaly.

[0089] The identity trust level, device trust level, network risk level, and abnormal behavior level are organized into a set of fuzzy context variables. The set of fuzzy context variables includes four fuzzy context variables, which correspond to the identity trust level, device trust level, network risk level, and abnormal behavior level, respectively. Each fuzzy context variable is represented in the form of a fuzzy membership vector consisting of multiple level labels and their corresponding membership degrees.

[0090] This invention constructs a fuzzy context variable set by performing fuzzification processing on multiple attributes in the context information set, including identity trustworthiness, device trustworthiness, network risk level, and behavioral anomaly level. Each variable uses a preset fuzzy membership function to calculate its fuzziness level and corresponding membership degree, forming a quantifiable fuzzy membership degree vector representation. This processing method enhances the system's ability to express uncertain information, enabling access control systems to make flexible judgments in complex or ambiguous scenarios, effectively improving the adaptability and fine-grained control level of risk identification, and overcoming the limitations of traditional rule-based mechanisms in handling fuzzy states.

[0091] In this embodiment, the construction of the improved fuzzy Bayesian network model to form a context-adaptive inference network specifically includes:

[0092] An improved fuzzy Bayesian network model is constructed, which includes a node set, a structural connection set, and a fuzzy conditional probability table set.

[0093] The node set takes a set of fuzzy context variables as input, and each node represents a fuzzy context variable. Node types include identity trust nodes, device trust nodes, network risk nodes, and abnormal behavior nodes.

[0094] Based on a pre-defined set of causal dependencies E = {(v i →v j )∣v i ,v j ∈V}, construct a structural connection set, which consists of multiple directed edges from the node set, forming a directed acyclic graph structure, wherein each directed edge v i →v j Represents the fuzzy context variable v i For fuzzy context variable v j The direct parent node, the connecting edge is used to establish fuzzy conditional dependency paths between the fuzzy context variables;

[0095] A time-sensitive fuzzy node structure is introduced, wherein the time-sensitive fuzzy node performs time-weighted processing on the historical state sequence and introduces a time decay factor on the fuzzy membership value of the node at the current time.

[0096] A context structure adaptation mechanism is constructed to dynamically adjust the set of structural connections based on the context state carried by the current access request, activating or blocking some edge connections to form an inference network structure.

[0097] Configure a fuzzy conditional probability table for a node, wherein the fuzzy conditional probability table defines the fuzzy output probability distribution of a child node under different combinations of parent node states;

[0098] The set of nodes, the set of structural connections, and the set of fuzzy conditional probability tables are integrated to construct a context-adaptive inference network.

[0099] This invention constructs an improved fuzzy Bayesian network model, using fuzzy context variables as node inputs and combining them with pre-defined causal dependencies to establish a directed acyclic graph structure, forming a context-adaptive inference network. The model introduces a time-sensitive fuzzy node structure, applying time-weighted processing to historical states to achieve dynamic perception of context evolution trends. Simultaneously, the system can adjust structural connections based on the current access request status, dynamically activating or blocking some causal paths, enhancing the inference network's environmental adaptability. By configuring fuzzy conditional probability tables, the flexible inference capability of risk assessment is improved, enhancing the access control system's accuracy and response flexibility in identifying dynamic risks in complex scenarios.

[0100] In this embodiment, the introduction of the time-sensitive fuzzy node structure specifically includes:

[0101] A time-sensitive mechanism is introduced into each fuzzy node in the improved fuzzy Bayesian network model to construct a historical state sequence, which consists of multiple context observation times.

[0102] At each time step t i For each fuzzy context variable in the fuzzy context variable set, calculate the fuzzy membership value to obtain the fuzzy state time series;

[0103] A time-weighting mechanism is introduced for fuzzy state time series, and each historical moment t is weighted according to its time position. i Assign weight values, which are calculated using an exponential decay method, meaning that the closer a time point is to the current moment, the higher its weight, and the earlier a historical time point, the lower its weight.

[0104] The fuzzy states of each fuzzy context variable are aggregated based on the time-weighted function to obtain the time-weighted fuzzy membership values:

[0105]

[0106] in, For fuzzy context variable v j The time-weighted fuzzy membership value at the current moment, k represents the historical state sequence. The time window length, where i represents the historical state sequence. The time index variable, t i For the i-th historical time point, w(t) i (t) represents time point t i The corresponding weight value, Fuzzy context variable v j At time ti The fuzzy membership value.

[0107] This invention introduces a time-sensitive fuzzy node structure into a fuzzy Bayesian network, constructs a fuzzy state time series using historical context states, and assigns weights to different time points based on a time-weighted mechanism to achieve dynamic perception of the trend of variable state changes over time. An exponential decay strategy is used to allocate weights, making the model more sensitive to state changes close to the current moment, thereby improving the ability to identify the evolution of potential risk behaviors. The fuzzy membership values ​​generated through time-weighted aggregation enhance the time-series perception capability of access risk assessment, effectively improving the system's judgment accuracy and response foresight in cases of continuous abnormal access or covert behavior.

[0108] In this embodiment, the construction context structure adaptation mechanism specifically includes:

[0109] Receive the context state vector carried by the current access request, wherein the context state vector is composed of the time-weighted fuzzy membership values ​​of each fuzzy context variable in the fuzzy context variable set;

[0110] The directed edges in the structure connection set are matched and determined based on a preset set of structure activation rules. The structure activation rules indicate that when a certain set of context state conditions are met, the corresponding connection edge is activated or blocked.

[0111] For each directed edge v i →v j Calculate the activation factor δ ij Based on the matching results in the set of structural activation rules, it is determined whether to activate directed edges. If the current context state satisfies a certain activation rule, it is marked as an activated state, denoted as δ. ij =1, otherwise it is marked as a shielded state, denoted as δ. ij =0;

[0112] Construct a dynamic structure connection set E ′ ={(v i →v j )∈ε|δ ij =1} serves as the reasoning network structure in the current context state, and the structure set E ′ It will adjust in real time according to changes in the access environment to form a context-adaptive inference network structure.

[0113] This invention achieves dynamic adjustment of the inference network structure based on the access context state by constructing a context structure adaptation mechanism. The system performs rule matching and activation determination on directed edges in the structural connection set based on the context state vector carried by the current access request, flexibly adjusting the inference path. When the access environment changes, the inference network can adjust the connection structure in real time according to the activation factor, retaining only dependencies related to the current risk, thus improving inference efficiency and judgment accuracy.

[0114] In this embodiment, the step of performing fuzzy inference and outputting the access risk level assessment result specifically includes:

[0115] The fuzzy context variable set is used as the input node variable and input into the context adaptive inference network.

[0116] Based on the currently active dynamic structure connection set, perform graph traversal reasoning, combine the fuzzy conditional probability table of each node, and update the fuzzy membership degree of the child node state according to the combination of the parent node state membership degree;

[0117] A fuzzy level label set is generated at the inference terminal node. The fuzzy level label set includes low risk level, medium risk level and high risk level. The fuzzy membership degree value of each fuzzy level label is calculated to form a membership degree distribution set. The membership degree distribution set includes the membership degree value of low risk level, the membership value of medium risk level and the membership value of high risk level.

[0118] The fuzzy level label set and the membership degree distribution set are combined to form the access risk level assessment result.

[0119] This invention performs fuzzy inference operations, dynamically updating the fuzzy membership values ​​of each node in a context-adaptive inference network based on activated structural connections and fuzzy conditional probability tables, and generating risk level labels and their membership distributions at terminal nodes. This mechanism enables fine-grained judgment of access request risks, supports simultaneous evaluation of memberships at multiple risk levels (low, medium, and high), improves the expressive power and discrimination accuracy of risk assessment results, and enhances the inference accuracy and decision-making rationality of access control systems in complex contexts.

[0120] In this embodiment, the generation of access control instructions through the access control decision mapping module specifically includes:

[0121] Receive access risk level assessment results, which consist of a set of fuzzy level labels and a corresponding set of membership degree distributions;

[0122] Based on the membership distribution set, an access risk level judgment operation is performed. When the membership value of a high-risk level is greater than the membership values ​​of medium-risk and low-risk levels, the access request is determined to be of high risk level, and an "access denied" operation instruction is generated.

[0123] When the membership value of the medium-risk level is greater than that of the high-risk and low-risk levels, the access request is determined to be of medium-risk level, and an operation instruction of "restrict access scope" or "perform secondary verification" is generated.

[0124] When the membership value of a low-risk level is greater than that of a high-risk or medium-risk level, the access request is determined to belong to the low-risk level, and an "access allowed" operation instruction is generated.

[0125] The control action corresponding to the access risk level judgment is used as the access control decision result, and an access control instruction is encapsulated and generated. The access control instruction includes at least one of allowing access, performing secondary verification, restricting the access scope, or denying access.

[0126] This invention, based on access risk level assessment results and membership distribution, determines the risk level and generates corresponding access control instructions. The system can automatically match control policies according to different risk levels, achieving tiered responses to access requests, including allowing access, restricting access scope, requiring secondary verification, or denying access. This mechanism enhances the flexibility and precision of access control, effectively avoiding over-authorization or false denial, and improving the system's access security capabilities and policy execution efficiency in dynamic environments.

[0127] In this embodiment, the access processing result includes control action result, trigger behavior record and context snapshot information; the control action result represents the response status after the access control command is executed; the trigger behavior record includes the record of key behavioral events triggered during the access process and their execution status; the context snapshot information saves structured data of user identity attributes, device attributes, network environment attributes, operation behavior characteristics and time and location attributes corresponding to the execution of the access control action.

[0128] This invention achieves full-process traceability and risk behavior logging by recording the control action status, triggering events, and context snapshot information in the access processing results. The structured storage of multi-dimensional context data at the moment of access facilitates subsequent audit analysis and strategy optimization, enhancing the system's security, controllability, and behavioral visualization capabilities.

[0129] In this embodiment, the context-adaptive inference network specifically includes:

[0130] Construct a set of policy feedback information, where each policy feedback information is composed of the corresponding access risk level assessment result and access processing result;

[0131] Based on the policy feedback information set, update and improve the fuzzy conditional probability table set in the fuzzy Bayesian network model;

[0132] Based on the policy feedback information set, the fuzzy membership function set is adjusted, and the fuzzy level boundary value is offset and corrected according to the actual feedback during the update process.

[0133] By introducing context snapshot information from the strategy feedback information, a structural dynamic adjustment index function is constructed. The structural dynamic adjustment index function enables or disables some directed connection edges in the inference network based on the state combination of the current fuzzy context variable set, and generates a new structural connection set.

[0134] Based on the updated fuzzy conditional probability table, fuzzy membership function set, and structural connection set, an updated context-adaptive inference network is constructed to realize the adaptive evolution of access control policies under different context states.

[0135] This invention achieves dynamic updates to fuzzy conditional probability tables, fuzzy membership functions, and structural connectivity relationships by constructing a policy feedback information set. Based on joint feedback of access results and context states, the system optimizes risk inference paths and membership degree partitioning, enhancing the model's adaptability to the evolution of access behavior. Simultaneously, a structural dynamic adjustment index function is introduced, enabling the inference network to adjust causal structural connections according to real-time context changes, achieving self-learning and continuous evolution of access control policies, and enhancing the system's stability and judgment accuracy in variable environments.

[0136] Example 1:

[0137] To verify the feasibility of this invention in practice, it was applied to a secure access control scenario within a large enterprise's internal information system. This enterprise has a large number of distributed office terminals, and employee identities, device access status, network environment, and operational behaviors exhibit highly dynamic and diverse characteristics. In previous access control strategies, the system primarily relied on static role configuration and single identity authentication, which could not effectively identify potentially risky access behaviors. Especially when facing login requests from untrusted devices, abnormal operation paths, or cross-regional login requests, authorization errors, response delays, and attack vulnerabilities arose.

[0138] In this test scenario, the system integrates the zero-trust model-based security access control method proposed in this invention, deployed at the enterprise's unified access gateway, for dynamic evaluation and control of each resource access request. The system first collects in real-time the user identity attributes, device status, network environment information, operational behavior characteristics, and access time and location carried in the access request, such as user authentication method, whether the device has trusted boot enabled, network address reputation, operational behavior sequence, and access geographical location. All information is then fuzzified to generate four types of fuzzy context variables: identity trust level, device trust level, network risk level, and abnormal behavior level.

[0139] Subsequently, these fuzzy variables are input into a time-sensitive fuzzy Bayesian network model constructed by this invention for risk level inference. This model can perform time-weighted aggregation based on historical context states and dynamically adjust the inference network structure according to the current state. For example, when the same user recently logs in to multiple high-risk addresses consecutively, and their actions exhibit abnormally high frequency of writes, the system identifies the evolution trend through historical state sequences, increases the weights of "network risk nodes" and "abnormal behavior nodes" in the inference path, thereby determining the risk level as "high" and triggering access restriction or secondary verification mechanisms.

[0140] Over a three-week testing period, the access control performance before and after deploying the method of this invention was compared as follows. The original system experienced a certain authorization delay when facing access requests from abnormal devices, with an average identification delay of 2.1 seconds. The system of this invention, relying on a fuzzy inference model, reduced this delay to 0.6 seconds. The false positive rate for abnormal access behavior decreased from 12.4% in the original system to 3.7%; the false negative rate for cross-regional login risk decreased from 9.8% to 1.2%. Regarding policy adaptability, the system of this invention continuously optimized the model structure through a policy feedback closed-loop mechanism, resulting in an average decision accuracy improvement of approximately 9.1% in the third week compared to the first week. The following are some access control evaluation and result comparison data.

[0141] Table 1 Evaluation of Access Control Effectiveness under Multidimensional Context-Driven Approach

[0142]

[0143] As can be seen from the table above, this invention demonstrates significant advantages in risk identification accuracy, control policy matching, and response timeliness in multi-dimensional context-driven access control scenarios. Firstly, regarding the accuracy of risk identification, the system can perform fuzzy reasoning by comprehensively considering identity trustworthiness, device trustworthiness, network risk level, and abnormal behavior level. For example, access requests numbered A1002 and A1008 simultaneously possess the characteristics of "medium or low identity trustworthiness," "untrusted or partially trusted devices," and "high network risk and abnormal behavior." The system correctly identifies these as high-risk behaviors and promptly denies access, effectively preventing potential unauthorized intrusion. Compared to existing systems, the false negative rate is significantly reduced, especially demonstrating stronger security sensitivity for cross-regional remote access (such as A1004) and high-frequency write operations (such as A1002).

[0144] Secondly, regarding the adaptability of access control policies, the system of this invention dynamically optimizes the inference network structure through a policy feedback mechanism. This allows it to adjust fuzzy conditional probabilities and structural connection edges based on access processing results during continuous operation, forming a closed-loop evolution process. The table shows that five access requests triggered policy updates after processing. Particularly after continuously identifying high-risk or potentially misjudged operations, the system can optimize risk mapping paths in real time, improving future judgment accuracy. This mechanism ensures the system can continuously learn and adapt to business changes, thereby enhancing long-term security control capabilities.

[0145] Finally, regarding the real-time performance of access responses, the system's average processing latency is controlled between 0.4 and 0.7 seconds, significantly better than the response latency of over 2 seconds typically found in traditional policy matching mechanisms. For example, comparing requests of different risk levels, A1001 and A1003, analysis and control decisions can be completed in less than 0.8 seconds, ensuring that business access is not significantly interrupted while providing high security. In summary, these results verify that this invention possesses excellent comprehensive performance in three key dimensions: risk identification accuracy, control policy evolution capability, and processing timeliness, and can meet the security protection needs of complex and dynamic access environments.

[0146] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A secure access control method based on a zero-trust model, characterized in that, Includes the following steps: Collect multi-source context information corresponding to access requests and generate a context information set; Based on a preset set of fuzzy membership functions, the set of context information is subjected to fuzzy encoding to obtain a set of fuzzy context variables. Construct an improved fuzzy Bayesian network model to form a context-adaptive inference network; The set of fuzzy context variables is input into the context adaptive inference network, fuzzy inference operations are performed, and the access risk level assessment result is output. Based on the access risk level assessment results, access control instructions are generated through the access control decision mapping module. Execute access response processing according to access control instructions and generate access processing results; The access risk level assessment results and access processing results are used together to construct policy feedback information and update the context-adaptive inference network.

2. The secure access control method based on a zero-trust model according to claim 1, characterized in that, The context information set includes user identity attribute information, device attribute information, network environment attribute information, operation behavior characteristic information, and time and location attribute information.

3. The secure access control method based on a zero-trust model according to claim 1, characterized in that, The process of performing fuzzy encoding on the context information set to obtain a fuzzy context variable set specifically includes: Based on a preset set of fuzzy membership functions, fuzzification processing is performed on various attribute information in the context information set. The set of fuzzy membership functions includes identity credibility function, device credibility function, network risk function, and behavior anomaly function. Input the user's identity attribute information into the identity credibility function to calculate the identity credibility level and the corresponding fuzzy membership degree; Input the device attribute information into the device credibility function to calculate the device credibility level and the corresponding fuzzy membership degree; Input the network environment attribute information into the network risk function to calculate the network risk level and the corresponding fuzzy membership degree; Input the operational behavior feature information into the behavior anomaly function to calculate the behavior anomaly level and the corresponding fuzzy membership degree; The identity trust level, device trust level, network risk level, and abnormal behavior level are organized into a set of fuzzy context variables.

4. The secure access control method based on a zero-trust model according to claim 1, characterized in that, The construction of the improved fuzzy Bayesian network model to form a context-adaptive inference network specifically includes: An improved fuzzy Bayesian network model is constructed, which includes a node set, a structural connection set, and a fuzzy conditional probability table set. The node set takes the set of fuzzy context variables as input, and each node represents a fuzzy context variable. Based on a preset set of causal dependencies, a structural connection set is constructed, which consists of multiple directed edges in the node set, forming a directed acyclic graph structure. A time-sensitive fuzzy node structure is introduced, wherein the time-sensitive fuzzy node performs time-weighted processing on the historical state sequence and introduces a time decay factor on the fuzzy membership value of the node at the current time. A context structure adaptation mechanism is constructed to dynamically adjust the set of structural connections based on the context state carried by the current access request, activating or blocking some edge connections to form an inference network structure. Configure a fuzzy conditional probability table for a node, wherein the fuzzy conditional probability table defines the fuzzy output probability distribution of a child node under different combinations of parent node states; The set of nodes, the set of structural connections, and the set of fuzzy conditional probability tables are integrated to construct a context-adaptive inference network.

5. A secure access control method based on a zero-trust model according to claim 4, characterized in that, The introduced time-sensitive fuzzy node structure specifically includes: A time-sensitive mechanism is introduced into each fuzzy node in the improved fuzzy Bayesian network model to construct a historical state sequence, which consists of multiple context observation times. For each fuzzy context variable in the fuzzy context variable set at each time step, the fuzzy membership value is calculated to obtain the fuzzy state time series; A time-weighting mechanism is introduced for fuzzy state time series, and each historical moment t is weighted according to its time position. i Assign weight values, which are calculated using an exponential decay method; The fuzzy states of each fuzzy context variable are aggregated based on the time-weighted function to obtain the time-weighted fuzzy membership values: in, For fuzzy context variable v j The time-weighted fuzzy membership value at the current moment, k represents the historical state sequence. The time window length, where i represents the historical state sequence. The time index variable, t i For the i-th historical time point, w(t) i (t) represents time point t i The corresponding weight value, Fuzzy context variable v j At time t i The fuzzy membership value.

6. A secure access control method based on a zero-trust model according to claim 4, characterized in that, The construction context structure adaptation mechanism specifically includes: Receive the context state vector carried by the current access request, wherein the context state vector is composed of the time-weighted fuzzy membership values ​​of each fuzzy context variable in the fuzzy context variable set; The directed edges in the structure connection set are matched and determined based on a preset set of structure activation rules. For each directed edge, determine whether to activate the directed edge based on the matching results in the set of structural activation rules. If the current context state satisfies a certain activation rule, mark it as active; otherwise, mark it as passive. A dynamic structural connection set is constructed, which consists of edge connections in the original structural connection set that satisfy the activation condition, and serves as the inference network structure in the current context state.

7. A secure access control method based on a zero-trust model according to claim 1, characterized in that, The specific steps of performing fuzzy inference operations and outputting access risk level assessment results include: The fuzzy context variable set is used as the input node variable and fed into the context adaptive inference network. Based on the currently active dynamic structure connection set, perform graph traversal reasoning, combine the fuzzy conditional probability table of each node, and update the fuzzy membership degree of the child node state according to the combination of the parent node state membership degree; A fuzzy level label set is generated at the inference terminal node. The fuzzy level label set includes low risk level, medium risk level and high risk level. The fuzzy membership degree value of each fuzzy level label is calculated to form a membership degree distribution set. The membership degree distribution set includes the membership degree value of low risk level, the membership value of medium risk level and the membership value of high risk level. The fuzzy level label set and the membership degree distribution set are combined to form the access risk level assessment result.

8. A secure access control method based on a zero-trust model according to claim 1, characterized in that, The specific steps of generating access control instructions through the access control decision mapping module include: Receive access risk level assessment results, which consist of a fuzzy level label set L and a corresponding membership degree distribution set μ. L constitute; Based on the membership distribution set, an access risk level judgment operation is performed. When the membership value of a high-risk level is greater than that of a medium-risk or low-risk level, the access request is determined to be of high risk level, and an "access denied" operation instruction is generated. When the membership value of the medium-risk level is greater than that of the high-risk and low-risk levels, the access request is determined to belong to the medium-risk level, and an operation instruction of "restrict access scope" or "perform secondary verification" is generated. When the membership value of a low-risk level is greater than that of a high-risk or medium-risk level, the access request is determined to belong to the low-risk level, and an "access allowed" operation instruction is generated. The control actions corresponding to the access risk level judgment are used as the access control decision results, and then encapsulated to generate access control instructions.

9. A secure access control method based on a zero-trust model according to claim 1, characterized in that, The access processing results include control action results, trigger behavior records, and context snapshot information; The control action result represents the response status after the access control command is executed; the trigger behavior record includes the record of key behavioral events triggered during the access process and their execution status; the context snapshot information saves structured data of user identity attributes, device attributes, network environment attributes, operation behavior characteristics and time and location attributes corresponding to the execution of the access control action.

10. A secure access control method based on a zero-trust model according to claim 1, characterized in that, The context-adaptive inference network specifically includes: Construct a set of policy feedback information, where each policy feedback information is composed of the corresponding access risk level assessment result and access processing result; Based on the policy feedback information set, update and improve the fuzzy conditional probability table set in the fuzzy Bayesian network model; Based on the policy feedback information set, the fuzzy membership function set is adjusted, and the fuzzy level boundary value is offset and corrected according to the actual feedback during the update process. By introducing context snapshot information from the strategy feedback information, a structural dynamic adjustment index function is constructed. The structural dynamic adjustment index function enables or disables some directed connection edges in the inference network based on the state combination of the current fuzzy context variable set, generating a new structural connection set. An updated context-adaptive inference network is constructed based on the updated fuzzy conditional probability table, fuzzy membership function set, and structural connection set.

Images