Blockchain-based surgical data privacy protection system

Abstract

The application relates to the technical field of medical data privacy protection, and discloses a surgical operation data privacy protection system based on a block chain. After a surgical operation data classification module obtains surgical operation data, the data content is analyzed, and data partitions are divided based on data attributes; a privacy policy mapping module obtains the data partitions, associates privacy policies stored in a block chain, and establishes a mapping relationship between the data partitions and the privacy policies; a key combination module extracts key partitions based on the mapping relationship, combines encryption keys, and calculates key matching degrees; a policy updating module updates the privacy policies and adjusts the mapping relationship based on the key combination and access records; and an access management module processes access requests and generates an access control library based on the updated mapping relationship. Through the collaborative operation of the multiple modules, the adaptability and reliability of surgical operation data privacy protection are improved.

Description

Technical Field

[0001] This invention relates to the field of medical data privacy protection technology, specifically a blockchain-based surgical data privacy protection system. Background Technology

[0002] In modern healthcare systems, surgical data contains a variety of sensitive information, including basic patient information, surgical plans, intraoperative monitoring data, and postoperative recovery records. This data is not only a crucial carrier of patient privacy but also a key basis for medical research and clinical decision-making. With the advancement of medical informatization, surgical data is often stored and managed using centralized databases. While this model brings convenience for data sharing, it also faces serious privacy and security challenges.

[0003] In centralized storage architectures, data management permissions are highly centralized. Once the database is attacked by hackers, internal personnel misconduct, or system vulnerabilities occur, it is extremely easy for a large amount of surgical data to be leaked. Among existing privacy protection technologies, traditional encryption methods mostly use a single key to encrypt all data, making it difficult to implement differentiated protection based on the sensitivity of different data. This results in some low-sensitivity data being over-encrypted, affecting usage efficiency, while high-sensitivity data is not adequately protected.

[0004] Another prominent issue is the lack of a dynamic adjustment mechanism for the formulation and implementation of privacy policies. Different types of surgical data correspond to different privacy requirements. For example, the privacy levels of patient identity information and surgical instrument usage records differ significantly, but the privacy policies of existing systems are often fixed and cannot be adaptively updated according to changes in data access scenarios, the identity of the accessing entity, and the purpose of the data. Furthermore, the lack of a precise mapping between privacy policies and data attributes leads to insufficient adaptability during policy implementation, either restricting legitimate data access needs or allowing unauthorized access.

[0005] Regarding data access control, existing systems rely heavily on static rules for access control, making it difficult to respond in real-time to changes in data privacy status. Key management mechanisms are fragmented, lacking targeted key combination strategies for critical data partitions, resulting in insufficient targeting and effectiveness of encryption protection. Furthermore, the utilization of access records is low, failing to provide effective basis for optimizing privacy strategies, leaving privacy protection in a passive defensive state and unable to cope with complex and ever-changing security threats. These problems severely restrict the secure sharing and compliant use of surgical data, necessitating a privacy protection solution capable of accurate data classification, dynamic policy adaptation, and intelligent key combination. Summary of the Invention

[0006] The purpose of this invention is to provide a blockchain-based surgical data privacy protection system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a blockchain-based surgical data privacy protection system, the system comprising:

[0008] The surgical data classification module is used to acquire surgical data, analyze data content, and divide data into partitions based on data attributes.

[0009] The privacy policy mapping module is used to obtain data partitions, associate them with the privacy policies stored on the blockchain, and establish a mapping relationship between data partitions and privacy policies.

[0010] The key combination module is used to extract key partitions based on mapping relationships, combine encryption keys, and calculate key matching degree.

[0011] The policy update module is used to update the privacy policy and adjust the mapping relationship based on key combinations and access records;

[0012] The access management module is used to process access requests and generate access control libraries based on the updated mapping relationships.

[0013] Preferably, the implementation methods of the surgical data classification module include:

[0014] For any data entry in the surgical data, obtain the attribute recognition model corresponding to the data entry; use the attribute recognition model to classify the data entry and obtain at least one attribute category; identify the keywords and related words existing in the data entries under the corresponding attribute category to form a text sample set; parse the keywords and related words in the text sample set respectively, and obtain the keyword area and related word area corresponding to the text sample set in sequence, and use them as data partitions.

[0015] Preferably, the implementation methods for obtaining the keyword area and related word area corresponding to the text sample set also include:

[0016] Keywords and related words in the text sample set are merged according to attribute categories to obtain a result of merging multiple attributes;

[0017] Extract keyword pairs from the attribute merging results and compare them with the keyword database to obtain the keyword region;

[0018] Extract the connection strength and distribution ratio of related words in the attribute merging results, and divide the attribute merging results according to the connection strength and distribution ratio of related words to obtain the related word region.

[0019] Preferably, the implementation methods for dividing data partitions also include:

[0020] Analyze the data partitions to obtain the access frequency and access time of the data partitions, fit the data partitions according to the access frequency and access time, and construct the mapping relationship between the data partitions and data attributes.

[0021] Preferably, the implementation of the privacy policy mapping module includes:

[0022] The privacy policy and access rules corresponding to the data partition are invoked to generate multiple unlabeled policy recognition results. The unlabeled policy recognition results represent privacy policies and access rules that are not associated with words in the data partition.

[0023] Determine whether multiple unlabeled policy identification results are target policy identification results. If they are target policy identification results, then treat the target policy identification results as privacy policies.

[0024] Preferably, the methods for establishing the mapping relationship between data partitions and privacy policies include:

[0025] Using the information from the keyword and related word regions within the data partitions, along with the descriptions and categories of the privacy policies, a mapping relationship is established between the data partitions and the privacy policies.

[0026] Preferably, the implementation of the key combination module includes:

[0027] Data partitioning and privacy policies are clustered according to data type, policy type, and policy function, and the largest cluster center after clustering analysis is set as the key partition.

[0028] Extract the feature terms of the key partitions, calculate the matching degree between each feature term, and set the common sequence related to the matching degree between each feature term;

[0029] By utilizing the common sequence related to the matching degree between each feature, the keys existing in the common sequence are extracted, and the longest matching sequence between each key is set. The length value of the longest matching sequence is set as the matching degree between each key.

[0030] The matching degree between each key is set according to the time distribution probability of each key, and the key combination is set.

[0031] Preferably, the implementation of the strategy update module includes:

[0032] Extract the temporal distribution probability of each key from the key combination;

[0033] Set the target path for the key combination according to the time period corresponding to the time distribution probability of each key.

[0034] The target paths of each key in the key combination are fitted to obtain the fitted target paths. The probability value of the fitted target paths in each time period is set as the combination probability of the key combination.

[0035] The probability of key combinations is compared with the references to access rules in the privacy policy. Differences are identified, and access rules in the privacy policy are categorized according to these differences to update the privacy policy.

[0036] Preferably, the access management module is implemented as follows:

[0037] Based on the mapping relationship between the updated data partitions and the privacy policy, the access targets in the access requests are extracted, and the access targets are sorted according to their probability of occurrence to obtain the access order in the access requests.

[0038] By combining access targets and access order in a structured manner, an access control library is obtained.

[0039] Preferably, the system further includes a permission allocation module, which is used to statistically analyze user role permissions and data access requirements based on the data partitioning and privacy policy mapping relationship, match permission availability with access requirements, calculate permission allocation difference, and generate permission allocation results;

[0040] The priority adjustment module is used to calculate the access task priority and task interval based on the permission allocation result, sort the permission call order, and generate an access task priority list.

[0041] The anomaly detection module is used to monitor changes in the data access environment based on the access task priority list, calculate access time deviations, and adjust permission allocation.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] This blockchain-based surgical data privacy protection system effectively addresses many pain points in traditional surgical data privacy protection through multi-module collaborative operation. The surgical data classification module, through in-depth analysis of data content and attribute-based partitioning, breaks away from the traditional "one-size-fits-all" approach in data management, enabling differentiated treatment of surgical data with varying levels of sensitivity, laying the foundation for subsequent precise protection. This partitioning method accurately identifies the privacy attributes of data, allowing for the separate management of highly sensitive patient identity information and core surgical plans from less sensitive routine operation records, avoiding the drawbacks of a single protection strategy.

[0044] The privacy policy mapping module associates data partitions with privacy policies stored on the blockchain. Leveraging the immutability of the blockchain, it ensures the authority and consistency of the privacy policies. Privacy policies stored on the blockchain can be verified by all nodes on the network, avoiding the risks of tampering or forgery that may occur in traditional centralized policy storage. Furthermore, by establishing a clear mapping relationship, each data partition can be matched with the most suitable privacy rules, making privacy protection measures more targeted.

[0045] The key combination module extracts key partitions and combines encryption keys based on mapping relationships, overcoming the limitations of traditional single-key encryption. By combining keys for key data partitions, encryption resources can be concentrated to strengthen the protection of highly sensitive data, while less sensitive data can be encrypted using relatively simplified methods, improving data utilization efficiency while ensuring core privacy and security. The calculation of key matching degree provides a reference for evaluating encryption effectiveness, ensuring that the key combination is compatible with the privacy requirements of the data partitions.

[0046] The policy update module dynamically adjusts the privacy policy based on key combinations and access records, enabling the privacy protection mechanism to adapt to changes in data access scenarios. By analyzing actual access patterns, it optimizes policy content and mapping relationships in a timely manner, avoiding issues of insufficient protection or excessive restrictions caused by policy rigidity, and ensuring that the privacy policy always remains in sync with data privacy needs.

[0047] The access management module processes access requests and generates an access control library based on the updated mapping relationship, enabling fine-grained management of access permissions. By applying dynamically adjusted privacy policies to the access control process, it can accurately identify legitimate access requests and block unauthorized requests. While ensuring data security, it also facilitates legitimate data use scenarios such as medical research and clinical collaboration, achieving a balance between data privacy protection and effective utilization. Attached Figure Description

[0048] Figure 1 This is a timing diagram of the blockchain-based surgical data privacy protection system described in this invention.

[0049] Figure 2 Workflow diagram for the surgical data classification module;

[0050] Figure 3 A flowchart of the privacy policy mapping module;

[0051] Figure 4 A flowchart for the strategy update module. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Please see Figure 1 This invention provides a blockchain-based surgical data privacy protection system, the system comprising:

[0054] This system achieves secure storage and access control of surgical data through a multi-layered data processing and privacy protection mechanism. The system consists of five core modules: a surgical data classification module responsible for attribute analysis and partitioning of raw surgical data; a privacy policy mapping module establishing the association between data partitions and blockchain-stored privacy policies; a key combination module generating dynamic encryption schemes based on the mapping relationship; a policy update module optimizing privacy policies according to access behavior; and an access management module executing real-time access control. These modules work collaboratively to form a closed-loop data privacy protection mechanism, with all operation records and policy updates stored immutably on the blockchain.

[0055] Example 1: See Figure 2 The surgical data classification module achieves refined partitioning of surgical data through multi-level attribute recognition and text analysis. This module first receives raw surgical data from medical information systems, including heterogeneous data sources such as structured electronic medical records, unstructured surgical report texts, and medical image metadata. For each input data entry, the system loads a pre-trained deep neural network model for initial classification. This model employs a multi-task learning architecture, with its input layer designed as a heterogeneous data adapter, automatically recognizing different data formats and converting them into a unified feature representation. The model's output layer generates attribute category labels including patient identification, surgical type, procedure steps, instrument usage, and complication records.

[0056] The text sample set is constructed using an extended method based on semantic association: the system performs word segmentation and part-of-speech tagging on the classified data entries, and uses an improved TF-IDF algorithm to extract keywords, where a semantic similarity correction factor is introduced in the weight calculation. For a given text segment... The keyword weight is calculated using the following formula:

[0057]

[0058] in: For keyword weight, For other terms, Maximum term frequency Indicates terms In the text Frequency of occurrence in The total number of documents in the corpus. For included terms The number of documents, Indicates terms Domain Dictionary The average semantic similarity is calculated using a pre-trained word vector model. The identification of related words employs a combination of dependency parsing and co-occurrence statistics. The system constructs a word co-occurrence network, calculates the mutual information values ​​between nodes, and retains edges above a threshold to form a set of related words.

[0059] The data partitioning process employs a two-level strategy: the system performs attribute category merging, mapping semantically similar categories to the same logical group. During merging, the system calculates the Jaccard similarity coefficient between categories and performs a union operation on categories with similarity exceeding a set threshold. Keyword extraction uses semantic clustering; the system maps keywords to a unified vector space, performs hierarchical clustering, and automatically determines the optimal number of clusters. The keywords at each cluster center constitute the core vocabulary of the keyword region. The system determines whether to include new keywords in the partition by calculating the cosine similarity between new keywords and the core vocabulary.

[0060] The partitioning of related word regions is based on a composite index of connection strength and distribution ratio. The system constructs a weighted undirected graph model to represent the relationships between related words, where nodes represent related words, and edge weights are determined by mutual information and context window statistics. Connection strength is calculated as the weighted degree centrality of nodes, while the distribution ratio is obtained by statistically analyzing the coefficient of variation of the frequency of related words in different document fragments. When determining partition boundaries, the system uses a spectral clustering algorithm to partition the related word graph, ensuring that related words within the same partition have high internal connection density and uniform distribution characteristics.

[0061] The dynamic optimization mechanism for data partitioning is centered on access behavior analysis. It records the access event sequence for each partition (extracting features such as timestamps, access subjects, and operation types), uses a variant of the Poisson process to model access frequency, introduces a periodic pattern detection algorithm to identify access time patterns, and continuously evaluates the performance of the partitioning scheme through an online learning framework. When metrics fail to meet standards, repartitioning is triggered. The system records the access event sequence for each partition, extracting features such as timestamps, access subjects, and operation types. Access frequency modeling uses a variant of the Poisson process; the system estimates the access intensity parameter per unit time and detects abnormal access patterns. Access time analysis introduces a periodic pattern detection algorithm to identify regular access characteristics such as daily and weekly cycles. Partition adjustment decisions are implemented through an online learning framework; the system continuously evaluates the information entropy and access efficiency metrics of the current partitioning scheme, triggering a repartitioning process when performance metrics fall below a threshold.

[0062] The partition mapping relationship is maintained using a version control mechanism. Each time the partition structure is adjusted, the system generates a new partition version identifier and stores the version change record on the blockchain. Old version partitions are retained for a limited time to support historical data access needs. The system automatically detects data overlap between different version partitions and performs reference counting management on duplicate data entries to ensure storage efficiency and data consistency.

[0063] Quality control of the text sample set is achieved through multi-dimensional verification. The system periodically performs sample sampling checks to evaluate keyword coverage and the connectivity of the related word network. For sample sets that fail to meet quality standards, the system initiates an automatic repair process, including measures such as supplementing an external medical terminology database and retraining the word vector model. All quality control records are written to the blockchain in the form of smart contracts, forming an immutable audit trail.

[0064] Secure isolation of data partitions is achieved through an encrypted storage scheme. A unique encryption key is assigned to each independent partition, and key management employs a threshold secret sharing scheme, with fragments stored across multiple blockchain nodes. During partition access, the system dynamically combines key fragments and performs decryption in memory, ensuring that the original data is not exposed to insecure environments. Key lifecycle management is strictly synchronized with partition versions; older version keys are automatically destroyed after version obsolescence.

[0065] Example 2: See Figure 3 The privacy policy mapping module establishes a precise mapping relationship between surgical data partitions and blockchain-stored privacy policies through a multi-stage matching and dynamic association mechanism. This module first retrieves the latest set of privacy policies from the blockchain distributed ledger. These policies are stored in the form of structured rules, including elements such as access control conditions, data operation permissions, and timeliness constraints. During system initialization, a policy parsing engine is loaded to convert the policy terms described in natural language into standardized machine-readable expressions, forming the basic framework of the policy rule base.

[0066] The initial matching of data partitions and privacy policies employs a semantic similarity calculation method. This involves extracting core features of the data partitions, including the term set in the keyword region and the semantic network structure in the related word region, and converting them into high-dimensional vector representations. Simultaneously, each policy in the policy rule base undergoes feature encoding to generate a corresponding policy description vector. During the matching process, the system calculates the cosine similarity between the partition vector and the policy vector, prioritizing candidate policies with similarity exceeding a preset threshold. For policy entries with insufficient matching, the system initiates a policy redirection mechanism, re-evaluating the policy's applicability by analyzing the logical structure of the policy rule and its correlation with the history of rejected policies.

[0067] The determination of unlabeled policy identification results adopts a rule-based reasoning method: the system maintains a policy knowledge graph, where nodes represent policy elements and edges represent logical relationships between elements. When encountering a policy not directly associated with the vocabulary of a data partition, the system performs multi-hop reasoning on the knowledge graph to find implicit semantic association paths. The reasoning process considers the contextual dependencies between policy elements, including the transitivity of time constraints and the inclusion relationship of permission conditions. For entries determined to be the target policy identification results, the system automatically generates semantic labels for them and establishes a weak association relationship with the corresponding data partition.

[0068] Dynamic optimization of the mapping relationship is achieved through a real-time feedback mechanism. The system monitors all data access events and records the deviation between actual access behavior and policy prediction results. When frequent policy matching failures are detected, the system triggers a mapping relationship re-evaluation process. This process analyzes pattern characteristics in historical access logs, identifies data usage scenarios not covered by the current policy, and generates policy adjustment suggestions. After administrator approval, the adjustment suggestions are updated to the blockchain network in the form of a smart contract to ensure that all nodes are synchronized with the latest mapping relationship.

[0069] The fine-grained association between data partitioning and privacy policies employs a multi-dimensional classification system. The system constructs a hierarchical classification framework, cross-categorizing data partitions according to dimensions such as sensitivity level, usage scenario, and data type. Privacy policies are classified in parallel based on dimensions such as protection strength, applicable objects, and timeframe. When establishing the mapping relationship, the system performs multi-dimensional matching to find the optimal intersection point within the classification system. For situations with multiple possibilities, the system uses a weighted voting mechanism to comprehensively consider the matching confidence of each dimension and select the optimal mapping scheme.

[0070] The detection and resolution of policy conflicts employs a logical consistency check method. The system periodically scans all active mapping relationships to detect any mutual exclusion between policy rules. Conflict detection is based on first-order logical reasoning, converting policy conditions into predicate expressions and identifying contradictions through resolution. When a conflict is detected, the system automatically selects the policy with higher priority based on its priority flag and creation timestamp, and generates a conflict resolution report for administrator review. All conflict handling records are written to the blockchain, forming a complete decision audit chain.

[0071] Version management of mapping relationships is achieved through the immutability of blockchain. Each time a mapping relationship is updated, the system generates a new version hash, containing a pointer to the previous version, the changed content, and a digital signature. The version history is organized in a Merkle tree structure, supporting rapid verification of the authenticity of any version. Old version mapping relationships are converted to cold storage after being phased out, but can still be traced back through a blockchain explorer. The version rollback mechanism allows the system to automatically revert to the previous stable version when abnormal mappings are detected.

[0072] The dynamic evaluation of strategy effectiveness employs a multi-indicator comprehensive evaluation system, continuously collecting execution data for each strategy, including operational metrics such as matching success rate, access interception rate, and strategy computational overhead. The evaluation model comprehensively considers the strategy's accuracy, coverage, and execution efficiency to generate a strategy effectiveness score. Low-scoring strategies are automatically placed in an optimization queue, where the system analyzes their performance bottlenecks and proposes targeted adjustment schemes. The effectiveness evaluation results serve as the basis for adjusting strategy weights, influencing the priority of strategy selection in subsequent mapping processes.

[0073] Cross-institutional sharing of privacy policies is achieved through a federated learning framework, allowing different medical institutions to participate in the collaborative training of policy models while maintaining the privacy of their local data. The system employs differential privacy technology to handle the transmission of policy parameters, ensuring that the original policy content cannot be inferred from the shared parameters. During federated training, incremental updates to the policy model of each node are verified and recorded through blockchain smart contracts, preventing malicious nodes from polluting the global model.

[0074] Example 3: The key combination module implements encrypted protection for surgical data partitions through multi-level clustering analysis and dynamic feature matching mechanisms. This module first receives the data partition set and its associated policy information from the privacy policy mapping module, performs standardized preprocessing on the input data, and unifies the data format and encoding method of partitions from different sources. The system uses an improved hierarchical clustering algorithm to group the data partitions and calculates the similarity matrix of each partition in three dimensions: data type, policy type, and policy function. During clustering, the system dynamically adjusts the distance threshold to ensure that partitions within each cluster have highly consistent privacy protection requirements.

[0075] Key partitions are determined using the silhouette coefficient evaluation method. This method calculates the ratio of the average distance between intervals within each candidate cluster to the average distance within each interval, forming a clustering quality index. When a cluster's quality exceeds a preset standard, its central partition is marked as a key partition. The system maintains a dynamically updated list of key partitions, reflecting changes in data access patterns in real time. The feature extraction process for key partitions employs a multi-granularity analysis method, capturing the essential characteristics of the partitions from three levels: structured attributes, text content, and access patterns.

[0076] The matching degree of feature items is calculated using an information gain-based method. The system constructs an association matrix between feature items to quantify the contribution of each feature to the partition security attributes. During the matching degree calculation, the system considers the conditional dependencies between feature items to avoid redundant calculations of correlations. The generation of public sequences uses a sliding window technique, where the system scans the occurrence patterns of all feature items and identifies frequently occurring feature combinations. The verification of public sequences is performed through distributed consensus via a blockchain network to ensure the authenticity and immutability of the sequences.

[0077] Key generation employs a feature-based hashing derivation method, assigning a unique identifier to each feature and generating initial key material through a cryptographic hash function. During key derivation, a dynamic time factor is introduced to impose time constraints on the generated keys. The key sequence is constructed using a binary tree structure, with each node containing partial key information and pointers to its child nodes. This structure supports efficient key retrieval and verification operations while maintaining the traceability of key relationships.

[0078] The key matching degree is calculated using the longest common subsequence algorithm. The system represents different keys as character sequences and constructs an edit distance matrix to find the optimal match. The matching degree is quantified using the following formula:

[0079]

[0080] in: For key matching degree, For the i-th key, For the first One key, Key and The longest common subsequence, Indicates the sequence length. Indicates the two keys within the time window The number of times they appear together within, This is a weighting adjustment factor used to balance structural similarity and temporal relevance. This formula ensures that the matching degree calculation considers both the static characteristics of the key and its dynamic usage patterns.

[0081] Key combination optimization employs a heuristic search algorithm. The system constructs a candidate space for key combinations and evaluates the security strength and performance overhead of each combination. During the search process, the system prioritizes key combinations with high matching degree and moderate usage frequency to avoid security risks caused by excessive key concentration. Combination optimization considers the current system's computational load and network conditions, dynamically adjusting the parallelism of key operations.

[0082] Key lifecycle management is implemented through blockchain smart contracts, with the generation, use, and destruction of each key recorded as transactions in a distributed ledger. The system periodically performs key health checks, assessing key usage frequency and security metrics. For keys nearing expiration, the system automatically initiates a key update process, generating a replacement key and smoothly migrating the data encryption scheme. Key revocation operations are broadcast to all network nodes via the blockchain consensus mechanism, ensuring timely effectiveness of revocation commands.

[0083] Access control for key combinations employs a multi-factor authentication mechanism, binding key usage permissions to the operation context, including factors such as request time, visitor identity, and operation type. Each key access attempt verifies whether the operation conforms to predefined policy rules and records detailed access audit logs. Abnormal access attempts trigger an automatic response mechanism, including temporary key locking and access source isolation protection measures.

[0084] The distributed storage of the key employs a secret sharing scheme, splitting the master key into shards and storing them across multiple blockchain nodes, with a reasonable reconstruction threshold set. The key recovery process requires a minimum number of shards to participate and verifies the requester's legitimacy using zero-knowledge proofs. The shard update mechanism ensures that even if some nodes are compromised, attackers cannot obtain the complete key information.

[0085] Performance optimization of the key combination module is achieved through a caching mechanism. The system maintains a hot key pool, storing frequently used key combinations. Cache management employs an adaptive replacement strategy, dynamically adjusting cache content based on the key's most recent usage time and frequency. Cold keys are stored in encrypted persistent media, using a lazy loading mechanism to reduce initialization overhead during system startup.

[0086] Example 4: See Figure 4 The policy update module dynamically optimizes the privacy policy through key usage analysis and access behavior modeling. This module continuously monitors the encryption operation records generated by the key combination module, extracting key activation time, usage frequency, and associated data partitions. The system establishes a probabilistic model of key time distribution, converting the occurrence time of each key into a time-series signal and identifying its periodic characteristics through spectral analysis. The target path for the key is constructed using a state transition network, where nodes represent key states, edges represent transition conditions between states, and network weights reflect transition probabilities. The system periodically executes a path optimization algorithm to eliminate redundant transitions in the network and maintain the simplicity of the path structure.

[0087] The probability of key combinations is calculated using a sliding window statistical method. The system maintains a dynamically adjusted time window, the size of which is automatically adjusted based on the stability of key usage patterns. Within the window, the system counts the occurrence frequency of each key combination and calculates its proportion relative to the total number of operations. An exponentially weighted moving average is used to smooth the probability values, reducing the impact of random fluctuations. The system compares the calculated results with the expected citation frequency recorded in the privacy policy to identify significant deviations from expectations.

[0088] The privacy policy discrepancy analysis employs a multi-dimensional evaluation framework, analyzing the shortcomings of the current policy from three levels: policy content, implementation effectiveness, and computational overhead. Content analysis checks the completeness and consistency of policy rules, ensuring no logical conflicts or coverage loopholes. Effectiveness evaluation is based on the degree to which actual access control decisions match expected decisions, quantifying the policy's execution accuracy. Overhead monitoring records the computational resource consumption of the policy matching process, identifying performance bottlenecks. The discrepancy identification process generates detailed discrepancy reports, marking the severity and scope of impact of each issue.

[0089] Policy classification and updates employ an incremental adjustment approach. Based on the discrepancy analysis results, policy issues are categorized into syntax errors, semantic deviations, performance defects, etc., and corresponding corrective measures are implemented for each type. Syntax errors are resolved by reconstructing the policy syntax tree; the system automatically detects and repairs structural problems in the rule expressions. Correction of semantic deviations requires referencing historical access data to adjust the thresholds or ranges of policy conditions. Performance optimization is achieved through reconstructing the policy matching algorithm, including introducing index structures and caching mechanisms for acceleration. All update operations are executed in a transactional manner to ensure the atomicity and consistency of policy changes.

[0090] The access management module implements priority-based data access control. The system parses input access requests and extracts core elements such as the request subject, target data, and operation type. The request verification process employs a multi-stage filtering mechanism, sequentially checking conditions such as the validity of identity credentials, the matching of operation permissions, and compliance with time constraints. The order of access targets is based on a dynamically calculated priority score, which comprehensively considers factors such as data sensitivity, request urgency, and user trust level. The system uses a weighted scoring model, with the weights of each factor dynamically adjusted according to the current system load and security posture.

[0091] The access control library is constructed using a structured storage scheme. The system features a two-layer index structure: the first layer is a hash-based fast lookup table that stores basic information and location pointers of the access target; the second layer is a detailed rule base based on a B+ tree, recording complete access control policies. Library structure optimization is achieved through periodic reorganization. The system analyzes changes in query patterns and adjusts the distribution and organization of the indexes accordingly. A version control mechanism ensures that updates to the access control library do not affect ongoing access operations, allowing for a smooth transition between old and new versions.

[0092] The detection and response to abnormal access employs a real-time monitoring mechanism. The system deploys lightweight detection agents at key nodes to collect temporal characteristics and behavioral patterns during the access process. The detection model is based on a hybrid architecture of rule engine and machine learning, incorporating both predefined security rules and automatically identifying new abnormal patterns. Response strategies are implemented in a tiered manner: low-risk anomalies trigger alerts and logging, while medium- and high-risk anomalies trigger automatic protective actions, including session termination and temporary account freezing.

[0093] Access auditing and traceability are achieved through the immutability of blockchain. Each access decision generates a corresponding audit record, containing key information such as the request content, decision basis, and execution result. Records are stored in encrypted form on a distributed ledger, organized using a Merkle tree structure, supporting rapid verification of the authenticity of any record. The audit query interface provides multi-condition filtering capabilities, allowing authorized personnel to retrieve historical access records by time range, operation type, user identity, and other dimensions.

[0094] The coordinated optimization of policies and access is achieved through a feedback loop. The system collects data on the actual effects of access control decisions and analyzes the degree of matching between policy rules and actual needs. The optimization suggestion generation module identifies frequently occurring policy exceptions and proposes targeted rule adjustment solutions. The policy testing environment allows new rules to be verified in an isolated sandbox before being deployed to the production system. All optimization operations form a complete change history, supporting issue traceability and policy rollback.

[0095] Example 5: The permission allocation module achieves refined control over surgical data access permissions through role-requirement matching and difference calculation. This module receives role definition information from the user management system and data requirement signals from the access management module, constructing a multi-dimensional permission allocation model. The role definition parsing process extracts features such as user job attributes, department affiliation, and years of service, mapping them to standardized role capability vectors. Data requirement analysis identifies elements such as target data type, operation level, and purpose of use from access requests, forming a requirement feature vector. The system maintains a dynamically updated permission matrix, recording the baseline values ​​of operation permissions for each role in different data partitions.

[0096] The permission difference calculation employs a vector space projection method, mapping the role's capability vector and required feature vector to a unified permission space, and calculating the distance between them in various dimensions. The distance metric considers the hierarchical structure of permissions, with higher-level permissions automatically including the access capabilities of lower-level permissions. The difference analysis process identifies gaps between a role's existing permissions and data requirements, distinguishing between mandatory core permissions and optional extended permissions. The system uses a constraint satisfaction algorithm to solve for the optimal allocation scheme, maximizing user access convenience while adhering to the principle of minimum necessary permissions.

[0097] The generation of permission allocation results considers multiple balancing factors. The system evaluates the security risks, operational efficiency, user experience, and other indicators of each candidate allocation scheme, and determines the final allocation decision through multi-objective optimization. The allocation process retains appropriate flexibility, allowing temporary elevation of permission levels in special circumstances, but this requires additional approval procedures and enhanced auditing. All allocation records are written to the blockchain in the form of smart contracts, forming an immutable history of permission changes. The permission revocation mechanism automatically detects situations where roles become invalid or their needs are terminated, and promptly revokes the relevant access permissions.

[0098] The priority adjustment module enables intelligent scheduling of access tasks, constructing a feature description for each pending access task, including attributes such as task initiation time, expected completion time, data sensitivity level, and user priority. The task sorting algorithm comprehensively considers the dynamic weights of these attributes to generate a globally optimal execution sequence. The weight adjustment mechanism responds to changes in system operating status, prioritizing efficiency indicators under high load and risk control during security alerts. Task interval optimization considers the balanced utilization of resources, avoiding local hotspots or prolonged idle periods.

[0099] The task priority list is maintained using a dynamic update strategy. The system monitors task execution progress and environmental changes in real time, making necessary adjustments to the list. Newly arrived high-priority tasks can be inserted into appropriate positions, while completed or canceled tasks are promptly removed from the list. Priority recalculation is triggered by events such as the release of new security policies, system resource fluctuations, and the detection of abnormal access patterns. List version management ensures the atomicity of adjustment operations, preventing consistency issues caused by concurrent modifications.

[0100] The anomaly detection module constructs a comprehensive security monitoring system, deploying a distributed sensor network to collect various signals in the data access environment, including network traffic patterns, user operation sequences, and system call logs. Environmental change detection employs time series analysis to identify abnormal fluctuations deviating from the normal baseline. Access time deviation calculation considers factors such as clock synchronization errors, network latency, and processing time, determining the acceptable deviation range through statistical process control methods.

[0101] Dynamic permission adjustments are based on risk assessment results, establishing a multi-level response mechanism to take corresponding measures for anomalies of varying severity. Minor anomalies trigger early warning notifications and enhanced verification; moderate anomalies initiate temporary permission downgrades and operational restrictions; and severe anomalies result in session termination and account isolation. Adjustment decisions consider business continuity requirements, maintaining a reasonable balance between security controls and availability. All adjustment operations generate detailed security incident reports for analysis and auditing purposes.

[0102] The federated learning framework supports the collaborative evolution of cross-institutional permission models, allowing healthcare institutions to share their experience and knowledge in permission management while maintaining local data privacy. The system employs encryption technology to protect the transmission of model parameters and verifies the authenticity of participating nodes through blockchain smart contracts. During collaborative training, the contribution of each node is evaluated through a consensus mechanism to prevent malicious nodes from polluting the global model. Model updates are verified through incremental deployment, and full implementation is only carried out after confirmation of effectiveness.

[0103] The access control audit trail employs blockchain-based evidence storage technology. Each access change operation generates a corresponding audit event, recording key information such as the operator's identity, the content of the change, and the basis for the decision. Event data is hashed and then written to a distributed ledger, forming a tamper-proof chain of evidence. The audit query interface supports complex condition searches, allowing analysis of access change history by time range, operation type, affected objects, and other dimensions. Alarm records from the anomaly detection module are stored in association with access operations, facilitating causal analysis of security events.

[0104] The human-machine collaborative interface for access control provides visual decision support, transforming complex permission relationships into intuitive graphical representations to help administrators understand the current permission distribution. The decision suggestion module provides alternative permission adjustment solutions based on historical data and similar cases. The manual intervention stage features a multi-level approval process to ensure that sensitive permission changes are fully justified and authorized. All interactions between manual operations and automated decisions are fully recorded, supporting post-event accountability.

[0105] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0106] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A blockchain-based surgical data privacy protection system, characterized in that, include: The surgical data classification module is used to acquire surgical data, analyze data content, and divide data into partitions based on data attributes. The privacy policy mapping module is used to obtain data partitions, associate them with the privacy policies stored on the blockchain, and establish a mapping relationship between data partitions and privacy policies. The key combination module is used to extract key partitions based on mapping relationships, combine encryption keys, and calculate key matching degree. The policy update module is used to update the privacy policy and adjust the mapping relationship based on key combinations and access records; The access management module is used to process access requests and generate an access control library based on the updated mapping relationship; The key combination module is implemented in the following ways: Data partitioning and privacy policies are clustered according to data type, policy type, and policy function, and the largest cluster center after clustering analysis is set as the key partition. Extract the feature terms of the key partitions, calculate the matching degree between each feature term, and set the common sequence related to the matching degree between each feature term; By utilizing the common sequence related to the matching degree between each feature, the keys existing in the common sequence are extracted, and the longest matching sequence between each key is set. The length value of the longest matching sequence is set as the matching degree between each key. The matching degree between each key is determined according to the time distribution probability of each key, and key combinations are set accordingly. The strategy update module is implemented in the following ways: Extract the temporal distribution probability of each key from the key combination; Set the target path for the key combination according to the time period corresponding to the time distribution probability of each key. The target paths of each key in the key combination are fitted to obtain the fitted target paths. The probability value of the fitted target paths in each time period is set as the combination probability of the key combination. The probability of key combinations is compared with the references to access rules in the privacy policy. Differences are identified, and access rules in the privacy policy are categorized according to these differences to update the privacy policy.

2. The blockchain-based surgical data privacy protection system as described in claim 1, characterized in that, The implementation methods of the surgical data classification module include: For any data entry in the surgical data, obtain the attribute recognition model corresponding to the data entry; use the attribute recognition model to classify the data entry and obtain at least one attribute category; identify the keywords and related words existing in the data entries under the corresponding attribute category to form a text sample set; parse the keywords and related words in the text sample set respectively, and obtain the keyword area and related word area corresponding to the text sample set in sequence, and use them as data partitions.

3. The blockchain-based surgical data privacy protection system as described in claim 2, characterized in that, Other methods for obtaining the keyword and related word regions corresponding to a text sample set include: Keywords and related words in the text sample set are merged according to attribute categories to obtain a result of merging multiple attributes; Extract keyword pairs from the attribute merging results and compare them with the keyword database to obtain the keyword region; Extract the connection strength and distribution ratio of related words in the attribute merging results, and divide the attribute merging results according to the connection strength and distribution ratio of related words to obtain the related word region.

4. The blockchain-based surgical data privacy protection system as described in claim 1, characterized in that, Other ways to implement data partitioning include: Analyze the data partitions to obtain the access frequency and access time of the data partitions, fit the data partitions according to the access frequency and access time, and construct the mapping relationship between the data partitions and data attributes.

5. The blockchain-based surgical data privacy protection system as described in claim 1, characterized in that, The implementation methods of the privacy policy mapping module include: The privacy policy and access rules corresponding to the data partition are invoked to generate multiple unlabeled policy recognition results. The unlabeled policy recognition results represent privacy policies and access rules that are not associated with words in the data partition. Determine whether multiple unlabeled policy identification results are target policy identification results. If they are target policy identification results, then treat the target policy identification results as privacy policies.

6. The blockchain-based surgical data privacy protection system as described in claim 3, characterized in that, The methods for establishing a mapping relationship between data partitions and privacy policies include: Using the information from the keyword and related word regions within the data partitions, along with the descriptions and categories of the privacy policies, a mapping relationship is established between the data partitions and the privacy policies.

7. The blockchain-based surgical data privacy protection system as described in claim 1, characterized in that, The access management module is implemented as follows: Based on the mapping relationship between the updated data partitions and the privacy policy, the access targets in the access requests are extracted, and the access targets are sorted according to their probability of occurrence to obtain the access order in the access requests. By combining access targets and access order in a structured manner, an access control library is obtained.

8. The blockchain-based surgical data privacy protection system as described in claim 1, characterized in that, It also includes a permission allocation module, which is used to calculate the user role permissions and data access requirements based on the data partitioning and privacy policy mapping relationship, match permission availability with access requirements, calculate permission allocation difference, and generate permission allocation results; The priority adjustment module is used to calculate the access task priority and task interval based on the permission allocation result, sort the permission call order, and generate an access task priority list. The anomaly detection module is used to monitor changes in the data access environment based on the access task priority list, calculate access time deviations, and adjust permission allocation.

Images