Digital archive management vulnerability assessment method and system

Abstract

The present application relates to the field of digital archives management and information security technology, specifically to a digital archives management vulnerability assessment method and system; containing data acquisition, semantic extraction, potential solution, entropy analysis and hierarchical defense module; the system uses a large language model to extract semantic vectors and construct a topology network; its core is to solve local semantic potential according to semantic coupling degree and edge reputation weight, obtain abnormal probability through Boltzmann distribution, and generate traceability computing power distribution coefficient based on Fermi-Dirac model; hierarchical defense is executed and weights are updated according to abnormal probability and topological entropy; the present application breaks through the limitations of traditional hash check, realizes the leap from byte physical detection to semantic logic level forgery identification, and ensures the logical integrity of archives.

Description

Technical Field

[0001] This invention relates to the field of digital archives management and information security technology, specifically to a method and system for assessing vulnerabilities in digital archives management. Background Technology

[0002] In the current digital archives management environment, massive amounts of unstructured archives face increasingly complex security threats, especially the risk of implicit semantic tampering that modifies the content logic. Existing security protection solutions generally adopt integrity verification mechanisms based on hash algorithms, verifying data consistency by comparing file digests. Although this solution can effectively detect physical changes at the file byte level, it lacks the ability to detect attacks that maintain the integrity of the file structure but only change the semantic logic, and cannot identify forgery behaviors such as referencing irrelevant clauses or forcibly associating with real archives. At the same time, if deep semantic tracing is performed on all archives at all times, it will generate huge computational overhead and response latency, resulting in wasted system resources in normal low-risk scenarios and scheduling lag during sudden attacks. Therefore, how to build a vulnerability assessment system that can accurately quantify logical conflicts, realize micro-level semantic perception and macro-level dynamic allocation of computing power, and has a self-evolving defense mechanism has become an urgent technical problem to be solved. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a method and system for assessing the vulnerability of digital archives management. Specifically, the technical solution of this invention is as follows:

[0004] A vulnerability assessment method for digital archive management includes:

[0005] Step 1: Collect archival data from the digital archive; extract high-dimensional semantic vectors from the archival data using a pre-set large language model; initialize and generate an archival topology network based on explicit reference relationships between archival data.

[0006] Step 2: Calculate the vector similarity between archive nodes based on high-dimensional semantic vectors; combine vector similarity with explicit reference relationships to determine semantic coupling degree;

[0007] Step 3: Combine semantic coupling degree with preset edge reputation weight to calculate the local semantic potential of the file node; input the local semantic potential into the preset Boltzmann distribution model to obtain the probability of abnormal existence of the file node;

[0008] Step 4: Calculate the topological entropy of the knowledge graph of the entire system based on the probability of anomaly existence; and generate the source tracing computing power allocation coefficient based on the topological entropy of the knowledge graph and the preset circuit breaker threshold using the preset Fermi-Dirac phase transition model.

[0009] Step 5: Based on the source tracing computing power allocation coefficient, perform hierarchical defense verification and output a source tracing confirmation signal;

[0010] Step 6: In response to the source tracing confirmation signal, update the edge reputation weight by combining the probability of anomaly existence with the semantic coupling degree and using preset error correction rules.

[0011] Preferably, semantic coupling is determined by combining vector similarity and explicit reference relationships, including:

[0012] Obtain explicit reference identifiers that indicate whether a reference chain exists between file nodes;

[0013] The semantic coupling degree is obtained by weighting and summing the vector similarity with the explicit reference identifier;

[0014] Semantic coupling degree is used to characterize the semantic rationality of reference relationships.

[0015] Preferably, the local semantic potential of the archive node is calculated as follows:

[0016] Get the set of neighboring nodes of the file node;

[0017] Calculate the square of the complement of the semantic coupling degree between the file node and each node in the set of neighboring nodes;

[0018] The local semantic potential is obtained by weighted summing of the square of the complement and the corresponding edge reputation weight.

[0019] Among them, local semantic potential is used to characterize the degree of logical conflict between the archive node and the citation environment.

[0020] Preferably, the probability of obtaining the existence of anomalies in file nodes includes:

[0021] Obtain the inverse temperature coefficient determined by statistical inversion of the system's historical false alarm rate;

[0022] Based on local semantic potential energy and inverse temperature coefficient, the probability of anomaly existence is determined by normalization using exponential and partition functions.

[0023] Among them, the probability of an anomaly is used to characterize the possibility that an archive node deviates from the trusted semantic system.

[0024] Preferably, the source tracing computing power allocation coefficients are generated using a preset Fermi-Dirac phase transition model, including:

[0025] Obtain the preset maximum computing power value and gain coefficient;

[0026] Calculate the difference between the topological entropy of the knowledge graph and the preset circuit breaker threshold;

[0027] Based on the product of the difference and the gain coefficient, a nonlinear mapping is performed between the exponential decay function and the maximum computing power value to generate the source computing power allocation coefficient.

[0028] Preferably, based on the source tracing computing power allocation coefficient, a tiered defense verification is performed, including:

[0029] If the knowledge graph topological entropy exceeds the preset circuit breaker threshold, the system is determined to be in normal daily condition, and the hash comparison algorithm is invoked for verification.

[0030] In response to the knowledge graph topological entropy being less than or equal to a preset circuit breaker threshold, the system is determined to be in a circuit breaker state, a deep semantic tracing algorithm is invoked for verification, and a tracing confirmation signal is output.

[0031] Preferably, the edge reputation weight is updated according to a preset error correction rule, including:

[0032] The difference between the source confirmation signal and the probability of an anomaly is calculated and used as the prediction error term;

[0033] Calculate the square of the complement of the semantic coupling degree as the semantic difference contribution term;

[0034] The prediction error term, the semantic difference contribution term, and the preset learning rate are multiplied to generate the weight correction value;

[0035] The weight adjustment value is then added to the current edge reputation weight to complete the update.

[0036] A vulnerability assessment system for digital archive management includes:

[0037] The data acquisition module is used to collect archival data and extract high-dimensional semantic vectors using a pre-set large language model;

[0038] The network building module is used to initialize and generate archive topology networks based on explicit reference relationships;

[0039] The potential energy calculation module is used to calculate the local semantic potential energy of archive nodes based on high-dimensional semantic vectors and edge reputation weights.

[0040] The entropy analysis module is used to calculate the topological entropy of the knowledge graph based on local semantic potential.

[0041] The decision response module is used to generate source tracing computing power allocation coefficients based on the knowledge graph topological entropy and to perform hierarchical defense verification.

[0042] The closed-loop feedback module is used to update the edge reputation weights based on the source tracing confirmation signal.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. This invention extracts high-dimensional semantic vectors from archival data through a large language model and calculates semantic coupling degree and local semantic potential by combining explicit citation relationships. This method breaks through the limitation of traditional hash verification that can only detect physical changes at the byte level. It can quantify the degree of conflict between archival nodes and citation environment from the semantic logic level, thereby accurately identifying logical forgery behaviors such as citing irrelevant clauses and forcibly associating with real archives, and ensuring the logical integrity of digital archive content.

[0045] 2. This invention introduces physical concepts and uses the Boltzmann distribution model to convert the local semantic potential energy of archive nodes into the probability of anomaly existence, and further calculates the topological entropy of the knowledge graph of the entire system. This mechanism maps the micro-logical conflicts of a single archive into a macro-level system disorder index, enabling the management system to quantitatively characterize the possibility of archive nodes deviating from the truth system, and providing accurate numerical basis for the assessment of security situation.

[0046] 3. This invention utilizes the Fermi-Dirac phase transition model to nonlinearly generate source tracing computing power allocation coefficients based on the difference between the knowledge graph topological entropy and the circuit breaker threshold, thus achieving hierarchical defense verification. Under normal conditions, the system only calls the low-energy hash comparison algorithm, but when an attack risk is detected, it can quickly exhibit a step response, mobilizing computing power to execute a deep semantic source tracing algorithm. This mechanism avoids the resource waste caused by 24 / 7 high-sensitivity operation, while ensuring timely response at the moment an attack occurs.

[0047] 4. This invention designs a closed-loop feedback module, which dynamically updates the edge reputation weights based on the difference between the source confirmation signal and the predicted probability through preset error correction rules. When a reference path is confirmed as the source of an attack, the system automatically adjusts the weights, so that the semantic potential generated by the path in subsequent calculations is significantly increased, thereby more sensitively capturing similar anomalies in the next detection. This mechanism realizes a closed loop throughout the entire process from data perception and risk quantification to dynamic defense and model updates. Attached Figure Description

[0048] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0049] Figure 1 This is a flowchart of the method of the present invention;

[0050] Figure 2 This is a structural diagram of the system of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0052] Example 1:

[0053] Please see Figure 1 A vulnerability assessment method for digital archive management includes:

[0054] Step 1: Collect archival data from the digital archive; extract high-dimensional semantic vectors from the archival data using a pre-set large language model; initialize and generate an archival topology network based on explicit reference relationships between archival data.

[0055] Step 2: Calculate the vector similarity between archive nodes based on high-dimensional semantic vectors; combine vector similarity with explicit reference relationships to determine semantic coupling degree;

[0056] Step 3: Combine semantic coupling degree with preset edge reputation weight to calculate the local semantic potential of the file node; input the local semantic potential into the preset Boltzmann distribution model to obtain the probability of abnormal existence of the file node;

[0057] Step 4: Calculate the topological entropy of the knowledge graph of the entire system based on the probability of anomaly existence; and generate the source tracing computing power allocation coefficient based on the topological entropy of the knowledge graph and the preset circuit breaker threshold using the preset Fermi-Dirac phase transition model.

[0058] Step 5: Based on the source tracing computing power allocation coefficient, perform hierarchical defense verification and output a source tracing confirmation signal;

[0059] Step 6: In response to the source tracing confirmation signal, update the edge reputation weight by combining the probability of anomaly existence with the semantic coupling degree and using preset error correction rules.

[0060] This embodiment provides a vulnerability assessment method for digital archives management. The method begins by collecting archive data from a digital archive repository. A pre-defined large language model is used to extract high-dimensional semantic vectors from the archive data. The large language model employs either BERT-based or RoBERTa, extracting the output of the last hidden layer as the high-dimensional semantic vector. This high-dimensional semantic vector is a numerical representation of unstructured text mapped to a multi-dimensional vector space using a deep learning model, enabling computers to quantify the semantic similarity between texts through geometric distance. An archive topology network is initialized based on explicit reference relationships between archive data. Vector similarity between archive nodes is calculated based on the high-dimensional semantic vectors, and semantic coupling is determined by combining vector similarity with explicit reference relationships. Finally, the semantic coupling is combined with pre-defined edge reputation weights to calculate... The local semantic potential of the archive node is used to obtain the probability of the archive node's abnormal existence by inputting the local semantic potential into a preset Boltzmann distribution model. The knowledge graph topological entropy of the entire system is calculated based on the probability of the abnormal existence. Based on the knowledge graph topological entropy and a preset circuit breaker threshold, a source tracing computing power allocation coefficient is generated using a preset Fermi-Dirac phase transition model. The source tracing computing power allocation coefficient is used to perform hierarchical defense verification and output a source tracing confirmation signal. In response to the source tracing confirmation signal, the edge reputation weight is updated by combining the probability of the abnormal existence and the semantic coupling degree through a preset error correction rule. This embodiment constructs a closed-loop physical model that integrates microscopic semantic potential, macroscopic topological entropy and dynamic immune response, realizing the effective identification of implicit semantic tampering in digital archives and filling the technical gap that traditional hash verification cannot cope with content logic tampering.

[0061] Example 2:

[0062] By combining vector similarity with explicit reference relationships, semantic coupling is determined, including:

[0063] Obtain explicit reference identifiers that indicate whether a reference chain exists between file nodes;

[0064] The semantic coupling degree is obtained by weighting and summing the vector similarity with the explicit reference identifier;

[0065] Semantic coupling degree is used to characterize the semantic rationality of reference relationships.

[0066] This embodiment provides a specific way to determine semantic coupling degree; semantic coupling degree Defined as a physical quantity that measures the semantic validity of reference relationships between nodes; obtains explicit reference identifiers that characterize whether a reference chain exists between archive nodes. When node With nodes When explicit reference chains exist The value is 1 if true and 0 otherwise; the semantic coupling degree is obtained by weighted summation of vector similarity and explicit reference identifier; the specific calculation follows the formula:

[0067] ;

[0068] It should be noted that this is introduced... This is to balance explicit references and potential semantic relationships, when there are no explicit references between nodes but the potential coupling degree needs to be calculated. Setting it to 0 automatically degenerates the formula into a pure vector similarity calculation; where, if If the calculation result is less than 0, set it to 0 to ensure... The value range is from 0 to 1. and The normalized high-dimensional semantic vector of the archive; The preset semantic weight coefficient is 0.6, derived from regression analysis of historical sample data. By introducing a weighted summation mechanism, the relevance of the content is carefully considered, effectively identifying forgery methods that cite irrelevant clauses or force connections, ensuring that only connections with both citation relationships and tight logical content have high coupling.

[0069] Example 3:

[0070] Solving the local semantic potential of archive nodes includes:

[0071] Get the set of neighboring nodes of the file node;

[0072] Calculate the square of the complement of the semantic coupling degree between the file node and each node in the set of neighboring nodes;

[0073] The local semantic potential is obtained by weighted summing of the square of the complement and the corresponding edge reputation weight.

[0074] Among them, local semantic potential is used to characterize the degree of logical conflict between the archive node and the citation environment.

[0075] This embodiment is a concretization of the calculation of the local semantic potential of an archive node; local semantic potential Defined as describing the logical repulsion force experienced by a single archive node in the current network environment; obtaining archive nodes. The set of neighboring nodes ; Calculate the square of the complement of the semantic coupling degree between the file node and each node in the set of neighboring nodes. The local semantic potential is obtained by weighting the square of the complement and the corresponding edge reputation weight; the specific calculation follows the formula:

[0076] ;

[0077] in, To prevent smooth terms with a denominator of zero, and to ensure that the local semantic potential of a node is constant when the node is isolated. Automatically converges to zero to avoid computational overflow; For the set of neighboring nodes The number of nodes in; The edge reputation weight is initially set to 1; the system can sensitively detect logical conflicts between a node and its citation environment. If a forged document cites a large number of real documents unrelated to its content, its coupling degree will be high. A lower value leads to an increase in semantic conflict terms, ultimately reducing the node's potential energy. Significantly increased.

[0078] Example 4:

[0079] Obtain the probability of an anomaly existing in the file node, including:

[0080] Obtain the inverse temperature coefficient determined by statistical inversion of the system's historical false alarm rate;

[0081] Based on local semantic potential energy and inverse temperature coefficient, the probability of anomaly existence is determined by normalization using exponential and partition functions.

[0082] Among them, the probability of an anomaly is used to characterize the possibility that an archive node deviates from the trusted semantic system.

[0083] This embodiment is a concretization of obtaining the probability of anomalies in archive nodes; it obtains the inverse temperature coefficient determined based on the statistical inversion of the system's historical false negative rate. ; This represents the system's detection sensitivity, and its value is derived from statistical inversion of the system's historical false negative rate data. A higher false negative rate indicates a lower sensitivity. The larger the set value, the greater the impact of small potential energy differences on the probability distribution; based on local semantic potential energy and inverse temperature coefficient, the probability of anomaly existence is determined through normalization using exponential and partition functions; the specific calculation follows the formula. , where the partition function Used for network-wide probability normalization, the calculation follows the formula. ,in, This represents the total number of archive nodes in the digital archive. This serves as an index for traversing all nodes in the network.

[0084] Inverse temperature coefficient The introduction of this feature enables the system to have macroscopic adjustment capabilities, and its numerical settings follow the formula. ,in, The basic sensitivity constant is set to a value range of [0.5, 1.5], with a typical empirical value of 1.0. It is used to define the baseline physical activity of the system when there is no history of missed reports. The system uses historical statistics on false negatives. In situations of severe security, even minor semantic conflicts can be amplified into a higher probability of anomalies, thus enabling dynamic control of system sensitivity.

[0085] Example 5:

[0086] Using a pre-defined Fermi-Dirac phase transition model, source tracing computing power allocation coefficients are generated, including:

[0087] Obtain the preset maximum computing power value and gain coefficient;

[0088] Calculate the difference between the topological entropy of the knowledge graph and the preset circuit breaker threshold;

[0089] Based on the product of the difference and the gain coefficient, a nonlinear mapping is performed between the exponential decay function and the maximum computing power value to generate the source computing power allocation coefficient.

[0090] This embodiment is a specific implementation of generating traceability computing power allocation coefficients; it obtains the preset maximum computing power value. With gain coefficient ; This represents the maximum limit of computing resources that the system can access. Define the steepness of the response curve, i.e., the acceleration of the control system switching from a steady state to a circuit breaker state; calculate the topological entropy of the knowledge graph. With preset circuit breaker threshold The difference; The calculation follows the formula:

[0091] ;

[0092] It should be noted that when obvious abnormal attack nodes appear in the system, the probability distribution of the anomaly will exhibit a spike-like shape, leading to an increase in the topological entropy of the knowledge graph. The value is significantly reduced, thus triggering the circuit breaker mechanism;

[0093] in, This represents the total number of archive nodes in the entire system. For the first The probability of anomalies in each node; this formula is used to quantify the overall disorder level of the current system; based on the product of the difference and the gain coefficient, a nonlinear mapping between the exponential decay function and the maximum computing power value is used to generate the source tracing computing power allocation coefficient; the specific calculation follows the formula:

[0094] ;

[0095] in, For the allocation coefficient of traceability computing power; The preset circuit breaker threshold is set to an empirical value of 0.7. This is the gain coefficient, with a value range of [5, 15]. In this embodiment, it is set to 10 to ensure that when the entropy value... When approaching the threshold, the computing power allocation coefficient can exhibit a steep step response characteristic;

[0096] The system utilizes the step characteristic of the Fermi-Dirac distribution to implement a phase-change response mechanism, ensuring low-energy operation during normal operation while rapidly mobilizing computing power during attacks, thus achieving a balance between performance and security. Considering the inherent latency in physical server resource scheduling, the system employs a preloading strategy to set time windows. Given a sampling period, calculate the ratio of the difference between the computing power allocation coefficients at the current time and the previous time. As an approximation of the first derivative, the rate of change is monitored in real time; when the rate of change of the computing power allocation coefficient exceeds a preset threshold, the standby computing cluster is preheated to approximate the instantaneous response characteristics of the theoretical model.

[0097] Example 6:

[0098] Based on the source tracing computing power allocation coefficient, perform tiered defense verification, including:

[0099] If the knowledge graph topological entropy exceeds the preset circuit breaker threshold, the system is determined to be in normal daily condition, and the hash comparison algorithm is invoked for verification.

[0100] In response to the knowledge graph topological entropy being less than or equal to a preset circuit breaker threshold, the system is determined to be in a circuit breaker state, a deep semantic tracing algorithm is invoked for verification, and a tracing confirmation signal is output.

[0101] This embodiment is a specific implementation of hierarchical defense verification; it responds to a knowledge graph topological entropy exceeding a preset circuit breaker threshold. The system is determined to be in normal daily operation, and a hash comparison algorithm is invoked for verification. The hash comparison algorithm primarily uses SHA-256 to verify file integrity, ensuring service efficiency. The system responds when the knowledge graph topological entropy is less than or equal to a preset circuit breaker threshold. If the system is determined to be in a circuit breaker state, a deep semantic tracing algorithm is invoked for verification, and a tracing confirmation signal is output. In this algorithm, 1 indicates tampering, and 0 indicates authenticity. The deep semantic source tracing algorithm includes cross-database citation verification and author style analysis. The specific steps of the author style analysis are as follows:

[0102] Extract the writing features of the target archive and construct a multidimensional style feature vector. The specific textual features include: frequency of function words, average sentence length, and punctuation density based on the N-gram algorithm; and the construction of a multi-dimensional style feature vector. The steps are as follows: First, the statistical values ​​of the above features are used to form an original feature set. Then, the data in the original feature set are standardized using Z-scores to eliminate dimensional differences. Next, L2 norm is used for normalization to generate a style feature vector for distance calculation. Finally, the historical archives of the author(s) are retrieved, and the average style centroid vector is calculated. If the named author is being recorded for the first time in the system and there is no historical archive data, then the global average style feature vector of all archives in the entire system will be retrieved. As an alternative benchmark, the covariance matrix of the style feature vectors is calculated based on this historical authentic archive set. and its inverse matrix This is used for subsequent Mahalanobis distance calculation;

[0103] calculate and Mahalanobis distance between ,like If the style dispersion threshold is exceeded, or if the citation source is found to be missing during cross-library citation verification, a source tracing confirmation signal will be output. ,otherwise The tiered defense strategy avoids the resource waste caused by the 24 / 7 high-sensitivity operation of conventional security systems and solves the drawback of low-sensitivity operation being unable to cope with sudden attacks.

[0104] Example 7:

[0105] The edge reputation weights are updated using preset error correction rules, including:

[0106] The difference between the source confirmation signal and the probability of an anomaly is calculated and used as the prediction error term;

[0107] Calculate the square of the complement of the semantic coupling degree as the semantic difference contribution term;

[0108] The prediction error term, the semantic difference contribution term, and the preset learning rate are multiplied to generate the weight correction value;

[0109] The weight adjustment value is then added to the current edge reputation weight to complete the update.

[0110] This embodiment is a concrete example of updating edge reputation weights using preset error correction rules; calculating the source tracing confirmation signal. Probability of anomaly The difference is used as the prediction error term; the square of the complement of the semantic coupling degree is calculated as the semantic difference contribution term; the prediction error term, the semantic difference contribution term, and the preset learning rate are then compared. Multiplication generates weighted adjustment values; A dimensionless constant is used to control the update step size, with a value range of [0.01, 0.1]. The weight correction value is added to the current edge reputation weight to complete the update. The specific calculation follows the formula:

[0111] ;

[0112] in, This is a preset upper limit for weights, used to prevent the weight values ​​from exploding and subsequent calculations from overflowing due to repeated accumulation during long-term system operation. For the updated edge reputation weight, The weights are as shown before the update; if a reference path has significant semantic differences and is ultimately confirmed to be the source of an attack, the system significantly increases the weight of that edge. This leads to the potential energy in the next calculation. It increases even more dramatically, achieving self-evolution of defensive capabilities.

[0113] Example 8:

[0114] Please see Figure 2 A vulnerability assessment system for digital archive management, comprising:

[0115] The data acquisition module is used to collect archival data and extract high-dimensional semantic vectors using a pre-set large language model;

[0116] The network building module is used to initialize and generate archive topology networks based on explicit reference relationships;

[0117] The potential energy calculation module is used to calculate the local semantic potential energy of archive nodes based on high-dimensional semantic vectors and edge reputation weights.

[0118] The entropy analysis module is used to calculate the topological entropy of the knowledge graph based on local semantic potential.

[0119] The decision response module is used to generate source tracing computing power allocation coefficients based on the knowledge graph topological entropy and to perform hierarchical defense verification.

[0120] The closed-loop feedback module is used to update the edge reputation weights based on the source tracing confirmation signal.

[0121] This embodiment provides a vulnerability assessment system for digital archive management. Logically, the system is divided into a data acquisition module, a network construction module, a potential energy calculation module, an entropy analysis module, a decision response module, and a closed-loop feedback module. The data acquisition module collects archive data and integrates a large language model interface to perform text-to-high-dimensional semantic vector conversion. The network construction module parses the reference format in the archive metadata and initializes and generates an archive topology network containing nodes and edges. The potential energy calculation module calculates data according to the formula... The system performs real-time calculations of the local semantic potential of each node; the entropy analysis module calculates node probabilities based on the Boltzmann distribution and aggregates them to calculate the topological entropy of the knowledge graph of the entire system; the decision response module incorporates a Fermi-Dirac phase transition model to calculate the computing power allocation coefficient in real time based on the entropy value and schedules hash validators or deep semantic tracing algorithms accordingly; the closed-loop feedback module receives the confirmation signal output by the decision response module and uses error correction rules to update the edge reputation weight table in the memory in reverse; the collaborative work between modules fully realizes the entire closed-loop process from data perception, risk quantification, dynamic defense to self-evolution, enhancing the digital archive system's ability to resist attacks by historical nihilism.

[0122] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for digital archive management vulnerability assessment, characterized by, The method comprises the following steps: Step 1, collecting archive data in a digital archive; Step 2, based on the high-dimensional semantic vector, calculating the vector similarity between the archive nodes; combining the vector similarity and the explicit citation relationship to determine the semantic coupling degree; Step 3, combining the semantic coupling degree and the preset edge reputation weight, solving the local semantic potential energy of the archive node; inputting the local semantic potential energy into the preset Boltzmann distribution model to obtain the abnormal existence probability of the archive node; Step 4, based on the abnormal existence probability, calculating the knowledge graph topology entropy of the whole system; based on the knowledge graph topology entropy and the preset fuse threshold, using the preset Fermi-Dirac phase transition model to generate the traceability computing power distribution coefficient; Step 5, according to the traceability computing power distribution coefficient, performing hierarchical defense verification, and outputting a traceability confirmation signal; Step 6, in response to the traceability confirmation signal, combining the abnormal existence probability and the semantic coupling degree, and updating the edge reputation weight through the preset error correction rule; Solving the local semantic potential energy of the archive node comprises: Obtaining the neighbor node set of the archive node; Calculating the complement square of the semantic coupling degree of each node in the neighbor node set of the archive node; Weighted sum of the complement square and the corresponding edge reputation weight to obtain the local semantic potential energy; Wherein, the local semantic potential energy is used to represent the logical conflict degree of the archive node and the citation environment. Combining the vector similarity and the explicit citation relationship to determine the semantic coupling degree comprises:

2. The method of claim 1, wherein, Obtaining the explicit citation identifier indicating whether there is a citation chain between the archive nodes; Weighted sum of the vector similarity and the explicit citation identifier to obtain the semantic coupling degree; Wherein, the semantic coupling degree is used to represent the semantic rationality of the citation relationship. Obtaining the abnormal existence probability of the archive node comprises:

3. The method of claim 1, wherein, Obtaining the inverse temperature coefficient determined based on the system historical false alarm rate statistics; Based on the local semantic potential energy and the inverse temperature coefficient, the abnormal existence probability is determined by exponential function and partition function normalization processing; Wherein, the abnormal existence probability is used to represent the possibility of the archive node deviating from the trusted semantic system. Using the preset Fermi-Dirac phase transition model to generate the traceability computing power distribution coefficient comprises:

4. The method of claim 1, wherein, Obtaining the preset maximum computing power value and gain coefficient; Calculating the difference between the knowledge graph topology entropy and the preset fuse threshold; Based on the product of the difference and the gain coefficient, the traceability computing power distribution coefficient is generated by nonlinear mapping through the exponential decay function and the maximum computing power value. According to the traceability computing power distribution coefficient, performing hierarchical defense verification comprises:

5. The method of claim 4, wherein, In response to the knowledge graph topology entropy being greater than the preset fuse threshold, it is determined that the system is in a normal state, and a hash comparison algorithm is called for verification; In response to the knowledge graph topology entropy being less than or equal to the preset fuse threshold, it is determined that the system is in a fuse state, a deep semantic traceability algorithm is called for verification, and a traceability confirmation signal is output. Updating the edge reputation weight through the preset error correction rule comprises:

6. The method of claim 1, wherein, Calculating the difference between the traceability confirmation signal and the abnormal existence probability as a prediction error term; Calculating the complement square of the semantic coupling degree as a semantic difference contribution term; ​ The prediction error term, the semantic difference contribution term and the preset learning rate are multiplied to generate a weight correction value; The weight correction value is superimposed on the current edge reputation weight to complete the update.

7. A digital archive management vulnerability assessment system for use in a digital archive management vulnerability assessment method according to any one of claims 1 to 6, characterized by It comprises: A data acquisition module is used to collect archive data and extract high-dimensional semantic vectors using a preset large language model; A network construction module is used to initialize the generation of an archive topology network based on explicit reference relationships; A potential calculation module is used to calculate the local semantic potential of the archive node based on the high-dimensional semantic vector and the edge reputation weight; An entropy analysis module is used to calculate the knowledge graph topology entropy based on the local semantic potential; A decision response module is used to generate a traceability computing power allocation coefficient based on the knowledge graph topology entropy and perform hierarchical defense verification; A closed-loop feedback module is used to update the edge reputation weight according to the traceability confirmation signal.

Images