Data full-life-cycle safety management method in swan gap system

Abstract

The invention provides a data full-life-cycle safety management method in a swan monk system, and the method comprises the steps: carrying out the standardized preprocessing and feature extraction of input data, extracting data features through a neural network model, carrying out the data classification, and generating a classification label and a sensitivity score; automatic distribution of authority is realized through an intelligent contract, and a preliminary authority strategy is generated; performing equipment identity verification and behavior analysis, dynamically evaluating an access risk by using a recurrent neural network model, adjusting an authority score according to the access risk, and generating a dynamic authority strategy and an audit record; performing compliance verification based on a compliance calculation formula, and generating a block chain verification result and a compliance notebook; and based on the block chain verification result, triggering a data destruction process through an intelligent contract, using a convolutional neural network to perform multi-coincidence rule verification, calculating a destruction threshold, executing data coverage clearing when a condition is satisfied, generating a destruction proof, and synchronizing the destruction proof to the block chain.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of data processing, and particularly relates to a data full life cycle security management method in a Hongmeng system. BACKGROUND

[0002] With the rapid development of information technology, the wide application of various intelligent devices and the Internet of Things, the generation and processing scale of data gradually become large, and data security has become one of the core problems in modern information systems. As a future-oriented distributed operating system, the Hongmeng system has been widely used in many fields, such as smart home, automobile, wearable device, etc. These devices have high distribution and heterogeneity, and the flow and storage of data between devices bring new security management challenges. In the Hongmeng system, data is not only stored and processed in the centralized server, but more often, the generation, storage, transmission, processing and destruction of data are distributed in different devices and network environments. This distributed data flow greatly increases the complexity of data management.

[0003] Most of the existing technologies at present are based on centralized system architecture for data security management, but this way is difficult to apply in distributed operating systems. The centralized data security management relies on a unified security control node, and such design leads to information asymmetry and unreasonable allocation of permissions in data access and management between multiple devices and nodes, which easily leads to risks such as data leakage, unauthorized access, malicious tampering, etc. In addition, most of the existing security management schemes adopt a static permission management model, which cannot dynamically adjust the access permissions of data according to the actual environment, and cannot effectively cope with the changing factors such as device environment, user behavior, data type, etc. For the management of the full life cycle of data, the existing technologies usually only focus on the encrypted storage and transmission of data, ignoring the comprehensive management of other key links in the data life cycle, such as data access control, permission management, compliance audit, etc., resulting in low management efficiency, and even difficult to meet the requirements of compliance and privacy protection.

[0004] Therefore, the existing technologies face a series of problems such as single point of failure of centralized control, limitations of static permission management, and incompleteness of data management when dealing with the full life cycle data security management in a distributed environment. SUMMARY

[0005] The purpose of the present application is to design a data full life cycle security management method in a Hongmeng system, which can realize a comprehensive and dynamic data security management mechanism in a distributed environment, especially on the basis of ensuring data privacy, compliance, transparency and non-tamperability, to realize efficient and secure management of the whole process of data life cycle.

[0006] In order to achieve the above object, the application provides a data full life cycle security management method in a Hongmeng system, which comprises the following steps:

[0007] The input data is standardized and preprocessed and feature extraction is performed, a neural network model is used to extract data features, data classification is performed based on an enhanced decision tree model and sensitivity calculation, classification labels and sensitivity scores are generated, and based on the classification results and distributed identity authentication, the automatic allocation of permissions is realized through a smart contract, and a preliminary permission policy is generated;

[0008] Based on the preliminary permission policy and the classification label, device identity authentication and behavior analysis are performed, a recurrent neural network model is used to dynamically evaluate access risk, and the permission score is adjusted according to the access risk, a dynamic permission policy and an audit record are generated;

[0009] The audit record and the dynamic permission policy are packaged as blockchain transaction data, stored through a decentralized consensus protocol, and compliance is verified based on a compliance calculation formula, a blockchain verification result and a compliance basis are generated;

[0010] Based on the blockchain verification result, a data destruction process is triggered through a smart contract, multiple compliance verifications are performed using a convolutional neural network, a destruction threshold is calculated, and data coverage is executed when the conditions are met, a destruction certificate is generated and synchronized to the blockchain.

[0011] Further, in the data classification, the sensitivity calculation includes a weighted combination of the basic score, the distributed factor and the noise estimation, wherein the distributed factor is calculated based on the hop count between devices.

[0012] Further, in the step of automatic allocation of permissions, the calculation of the permission score includes a weighted combination of the sensitivity score, the identity matching degree and the device security, and a context correction term based on device load is introduced.

[0013] Further, the recurrent neural network model is a four-layer LSTM structure; its input includes the frequency sequence of access requests, the timestamp sequence and the preliminary permission score; the network outputs a multi-dimensional behavior risk vector, which represents the frequency risk, the mode abnormal risk and the context interference risk.

[0014] Further, the adjustment of the permission score is based on the preliminary permission score, the behavior risk factor, the identity verification score and the classification label factor, and is realized through a nonlinear function combination.

[0015] Further, the audit record is generated in JSON format, including timestamp, visitor identity, permission change details, behavior risk data and classification label, and uses local encryption signature.

[0016] Further, the storage process through the decentralized consensus protocol adopts a practical Byzantine fault-tolerant protocol, and the number of fault-tolerant nodes is dynamically adjusted based on device load and audit record risk.

[0017] Further, in the compliance verification, the calculation of the compliance score is based on the dynamic permission score, the average value of the behavior risk, the block height, and the number of violation marks, and is evaluated through a nonlinear function.

[0018] Further, in the data destruction process, the calculation of the destruction threshold is based on the compliance score, the average value of the verification vector, the block hash entropy, and the chain length, and is adjusted through the inverse tangent function and the fourth root function.

[0019] Further, the data clearing operation includes multiple rounds of overwrite with random bytes and zero values, and generates a destruction hash as an unforgeable destruction proof.

[0020] The beneficial technical effects of the present application are at least the following points:

[0021] To solve the above problems, the present application provides a data full life cycle security governance method in a Hongmeng system, which introduces a decentralized technical architecture to avoid the single point failure problem of centralized systems, making the data governance more efficient and secure in a distributed environment. Secondly, by using smart contracts and dynamic permission adjustment mechanism, the access permission of data can be flexibly adjusted according to the type of data, the identity of the visitor, the security of the device, etc., and the governance strategy is dynamically updated according to the real-time environmental changes, solving the limitations of the existing static permission management. In addition, by embedding data audit and compliance check into every link of the data life cycle, it is ensured that all data operations can be transparently recorded and cannot be tampered with, improving the transparency and credibility of data governance. Finally, the method introduces CNN verification and threshold calculation in the destruction stage to ensure the complete removal of residual data, meeting the requirements of privacy regulations. Through the above mechanisms, the present application can provide an efficient and scalable security governance solution in a complex environment of multiple devices and multiple scenarios, effectively solving the defects in the current technology and meeting the increasingly stringent requirements of data privacy protection and compliance. BRIEF DESCRIPTION OF DRAWINGS

[0022] The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0023] Figure 1 The present application is a data full life cycle security governance method in a Hongmeng system. DETAILED DESCRIPTION

[0024] Embodiments of the present application are described below in the detailed description and illustrated in the accompanying drawings by which like or similar elements, or elements having the same or similar functions are designated with the same or similar reference numerals throughout the attached drawings and this disclosure. The embodiments described below are exemplary only, and are not to be construed as limiting the present application.

[0025] In one or more embodiments, as shown in Figure 1 A method for data life cycle security management in a Hongmeng system is disclosed, the method comprising the following:

[0026] S1: Standardized preprocessing and feature extraction are performed on the input data, a neural network model is used to extract data features, and data classification is performed based on an enhanced decision tree model and sensitivity calculation, to generate a classification label and a sensitivity score; based on the classification result and distributed identity authentication, the automatic allocation of permissions is realized through a smart contract, to generate a preliminary permission policy;

[0027] Specifically, first, the newly generated input data is preprocessed to extract and optimize the features, to adapt to the distributed characteristics of data flow in the Hongmeng system. The specific operation is to convert the original data value to the range of [0, 1] through a standardization process, for example, the heart rate reading sequence is divided by the physiological upper limit 200, to ensure the data compatibility between heterogeneous devices such as wearable wristbands and mobile phones. Subsequently, a three-layer feedforward neural network is used to realize feature extraction on the local AI engine of the Hongmeng device: the input layer receives the standardized data sequence, and the dimension is equal to the number of features, such as 5 health indicators including heart rate, blood pressure, etc.; the hidden layer uses the ReLU activation function, and contains 32 neurons to capture the nonlinear patterns in the data, such as sudden health fluctuations; the output layer generates a compressed feature vector. The network is fine-tuned by the local historical data of the device, and device context weights are added in the training to handle the noise of distributed data, such as health reading deviation caused by network delay, to ensure that the feature extraction is more robust, so as to better handle data variability in the multi-device environment of the Hongmeng system, improve the accuracy of classification and the reliability of subsequent permission allocation, reduce the delay compared to traditional centralized preprocessing, and improve the adaptability to heterogeneous sources.

[0028] Next, the extracted feature vector is used for data classification, which is realized through an enhanced decision tree model, which contains 5 decision nodes: the first node checks whether it contains personal information based on the existence of the user ID in the metadata; the subsequent nodes evaluate the real-time performance, generate a time difference less than 1 minute from the current time, and assess the privacy impact, the health indicators in the feature vector exceed the threshold of 0.5. In view of the distributed flow risk in the Hongmeng scenario, such as the risk of data leakage during transmission between insecure devices, the model introduces a sensitivity calculation formula, adding a distributed regularization term:

[0029] ;

[0030] wherein, is the sensitivity score, ranging [0,1], >0.6 as high sensitive label; is the base score, accumulated from decision tree nodes, such as 0.4 if personal information exists; is the distributed factor, hop count between devices, calculated for the Harmony Soft Bus, such as >2 to increase risk; is the noise estimation, based on feature vector variance; is the parameter. The formula simulates the risk of flow decay through the distributed regularization term and prevents noise over-amplification through the logarithmic noise term , which ensures more accurate classification for privacy governance, reduces 20% misclassification rate compared to general models, provides a more reliable classification basis for dynamic permission adjustment, reduces the possibility of data leakage in distributed environments, and improves the limitations of existing static classification on real-time variations.

[0031] After classification, the smart contract deployed on the Harmony distributed soft bus is activated, written in a language similar to Solidity, which first verifies the distributed identity: the contract calls the Harmony DistributedIdentity module, inputs the device public key and data signature, and continues if the elliptic curve encryption verification matches. Permission allocation is based on the sensitivity score of the classification label and identity matching, using the following weighted formula to calculate the preliminary permission:

[0032] ;

[0033] wherein, is the permission score, [0,1], threshold determines read / write / execute rules, such as >0.7 full permission; weights, w1=0.4 for sensitivity s1, w2=0.3 for identity matching s2, w3=0.3 for device security s3; Sigmoid term , μ=0.1, δ=0.5, introduces context correction c, based on device load in Harmony scenario, such as CPU >80% then c=-1 to reduce permission. This term adjusts the permission through context correction, targets device load fluctuations, ensures that the allocation adapts to real-time scenarios, improves the robustness of governance, and better handles variations in distributed environments than existing static permission models, improving overall system availability.

[0034] S2: Based on the preliminary permission policy and classification label, perform device identity verification and behavior analysis, use a recurrent neural network model to dynamically evaluate access risk, and adjust the permission score based on the access risk to generate a dynamic permission policy and audit record;

[0035] Specifically, this step realizes dynamic control of permissions and generation of audit records through device identity verification and real-time behavior analysis. This design targets the real-time variability of data access in the Hongmeng system, such as the risk caused by changes in user behavior across multiple devices, ensuring that permissions adapt to environmental changes while generating tamper-proof records to support subsequent blockchain audits, thereby bridging the gap between initial allocation and final verification, improving the adaptability and security of distributed data governance, and overcoming the limitations of existing static models. Through close linkage with Step 1, such as introducing sensitivity weighting in behavior analysis using classification labels, the continuous protection of the entire life cycle is strengthened, effectively preventing dynamic threats in Hongmeng ecosystems such as smart homes or automotive scenarios, such as abnormal access to health data under network fluctuations.

[0036] First, the input classification label and preliminary permission policy are subjected to device identity verification to confirm the legitimacy of the access request. The specific operation is to call the Hongmeng DistributedIdentity module, input the device public key of the visitor and the request signature, and perform elliptic curve encryption comparison with the identity matching in the preliminary permission policy. If the matching degree exceeds the 0.8 threshold, based on the public key hash similarity calculation, for example, using SHA-256 hash difference normalized to [0, 1], then continue, otherwise reject access and record a preliminary alarm. This verification process is performed on the local device to ensure efficient processing and avoid centralized delays, targeting the risk of identity forgery for distributed devices, such as external devices simulating access to health data in a smart home, leading to privacy leaks, and dynamically adjusting the threshold based on the classification label, such as increasing the threshold to 0.9 for high-sensitivity labels to increase the protection layer for privacy-intensive data, reducing the identity fraud rate by 15% compared to general verification methods in a distributed environment, and improving the response capability to real-time threats through label-driven threshold optimization.

[0037] Next, behavioral analysis is performed to dynamically adjust permissions using a four-layer recurrent neural network, RNN, implemented on the Harmony AI engine: the input layer receives sequential data, including current access behavior such as access frequency, timestamp sequences, and scores from the preliminary permission policy as initial hidden states to maintain continuity with Step 1; two hidden layers each contain 64 LSTM units to process temporal patterns such as 5-minute access sequences and capture anomalies such as sudden high-frequency read / write operations on health data; the output layer generates a behavior risk vector with a dimension of 3, representing frequency risk, pattern anomaly, and context interference. The network is fine-tuned using device local logs, embedding the inter-device delay factor of the Harmony bus query in the LSTM gate function as an additional input dimension to handle the distributed behavior peculiarities of Harmony, such as sequence discontinuity and network delay interference caused by cross-device access, ensuring more accurate analysis to identify malicious patterns, such as distinguishing normal sensor jitter from abnormal intrusion attempts in a car driving scenario, thereby providing more robust real-time responses, improving the limitations of existing static analysis of behavior, and amplifying the sensitivity of anomaly detection under high-sensitive classification labels, reducing the risk of data tampering in distributed flows, and improving the overall system's adaptability to variable environments.

[0038] Based on the behavior risk vector and classification label, the dynamic permission adjustment is calculated using the following formula to update the permission score:

[0039]

[0040] wherein, is the dynamic permission score, [0, 1], used to update read / write / execute rules, such as downgrading to read-only if <0.5; is the score of the preliminary permission policy, directly taken from the input; is the risk factor, averaged from the RNN output vector, such as >0.4 indicating high risk; is the verification score, calculated from identity matching, such as 0.9; is the label factor, with high-sensitive classification labels set to l=1.5, otherwise 1; is the parameter. This formula simulates the periodic fluctuations of sensitivity through the sine-square root label term and combines the square root regularization to prevent overfitting, addressing the special uncertainty of high-sensitive data in distributed flows, such as the intermittent risk of health data during multi-device synchronization, providing a nonlinear penalty mechanism, combined with the tanh verification term to ensure that dynamic control more accurately prevents real-time threats, and introducing oscillatory adjustments under high labels to simulate the network pulse effects of Harmony scenarios, improving real-time protection of health data, reducing the number of variables while strengthening depth, and being more adaptable to distributed variations than traditional Sigmoid. Through this adjustment mechanism, the limitations of existing static permissions are improved, and governance efficiency is enhanced. ​

[0041] After adjustment, generate audit record: compile a JSON structure, including timestamp, current Unix timestamp, visitor ID, extract from verification, permission score before and after update, compare preliminary and dynamic values, behavior risk details, serialize and quantize from RNN vector, such as risk average, classification label, copy from input to maintain consistency, ensure transparent record of each operation, and use local encryption signature of Harmony, such as key derivation based on preliminary permission policy, without external storage, but in block chain compatible hash chain format, support non-tamperable verification of subsequent steps, thereby strengthening end-to-end integrity of audit.

[0042] The whole process is executed on the access device, with synchronous update of the policy to the relevant nodes using the Harmony software bus, ensuring minimal delay, such as <50ms, and seamless continuity of input and output in a distributed environment, paving the way for compliance audit and destruction, highlighting the improvement over existing technologies for incomplete governance through this behavior-driven record generation.

[0043] The output is a dynamic permission policy, an updated structured object including read / write / execute permission scores and rules derived from dynamic permission score calculation, and an audit record, a JSON object including operation details derived from behavior analysis and adjustment compilation.

[0044] S3: Package the audit record and dynamic permission policy as block chain transaction data, store it through a decentralized consensus protocol, and verify compliance based on compliance calculation formula to generate block chain verification result and compliance certificate;

[0045] Specifically, this step stores and verifies compliance of the output of step 2 through a decentralized block chain mechanism, ensuring transparency and non-tamperability of data operations and supporting the final destruction step. This design addresses the audit challenges of distributed data in the Harmony system, such as the risk of tampering or loss of audit records between multiple devices, and uses the consensus protocol of the block chain to achieve cross-device verification, thereby connecting dynamic control and destruction, improving compliance and privacy protection throughout the life cycle, overcoming the single point of failure problem of existing centralized audits, and through close linkage with step 2, such as using Protobuf details of the audit record as block transaction payload and dynamic permission policy score as verification weight, strengthening end-to-end non-repudiation, effectively ensuring traceability of operations in Harmony ecosystems such as wearable device health data sharing scenarios, and providing a credible basis for destruction.

[0046] First, the input audit records and dynamic permission policy are packaged into blockchain transaction data. Specifically, the audit records of the Protobuf object are parsed to extract the timestamp, visitor ID, permission change, behavior risk details, and classification label, which are combined with the score and rules of the dynamic permission policy into a Merkle tree leaf node. The root hash is calculated layer by layer using SHA-256 hashing to ensure data integrity and tamper resistance. This packaging process broadcasts data to neighboring nodes through the distributed bus of the Hongmeng, forming a temporary consortium chain to avoid the bottleneck of centralized storage. The leaf nodes are dynamically sorted according to the score of the dynamic permission policy, for example, when the score is <0.6, they are placed near the tree root to be verified first, thereby addressing the risk of record fragmentation in distributed devices, such as inconsistent audit logs in multiple devices in a smart home, which can lead to compliance audit failures. This improves packaging efficiency, reducing network load by 25% compared to general methods, and better manages priority through a permission-driven tree structure, ensuring that high-risk records are promptly processed and improving the responsiveness of overall governance.

[0047] Next, consensus and storage are performed on the decentralized network using a practical Byzantine fault tolerance, PBFT, consensus protocol integrated with Hongmeng on device nodes: each node acts as a validator, inputting the hash of the packaged transaction and audit records; the protocol collects at least 2f+1 signatures through three-phase pre-preparation, preparation, and submission, where f is the number of fault-tolerant nodes, for example, f=3 for a network of 10 nodes. Nodes use local CPU computing to calculate signatures using the Elliptic Curve Digital Signature Algorithm, ECDSA, to ensure that blocks are appended to local chain copies after consensus is reached and synchronized to all nodes. This protocol dynamically adjusts f based on device load and risk details of audit records, embedding the node's computing power from soft bus queries as a signature weight factor during the preparation phase, ensuring that consensus more accurately adapts to malicious nodes or network partitions, such as favoring high-performance nodes in the automotive scenario to handle high-frequency audits of health data. This provides more robust storage for Hongmeng's distributed resource limitations, such as low-power wearable devices, improves the uniform fault tolerance limitations of existing PBFT, and increases the f value under low-privilege dynamic policies to reduce the risk of distributed tampering and improve the stability of the overall chain. It also better balances resource consumption and reliability compared to traditional consensus mechanisms.

[0048] Based on the consensus block, compliance verification is performed, and a compliance score is calculated using the following formula to evaluate each record:

[0049] ;

[0050] where, is the compliance score, [0,1], >0.7 indicates compliance; is the score of the dynamic permission policy, taken directly from the input; is a risk threshold, averaged from the behavior risk details of the audit record, e.g. 0.5; is a block height, from consensus generation, e.g. 10; is a violation flag, from the permission change count in the record, e.g. >2 is high, like 1; is a parameter. The formula preserves the cubic root-exponential tail The cubic root function simulates the asymptotically stable growth of the chain height, and the exponential tail combines the violation accumulation without penalty, providing a non-linear dynamic assessment for the long-term accumulated uncertainty of the audit records on the distributed chain, e.g. the gradual risk of the health data chain growth, ensuring that the verification more accurately identifies potential violations, and gradually becomes conservative at high block height to simulate the resource exhaustion impact of the Harmony network, improving the compliance protection of multi-device long-time data synchronization, reducing the number of variables while strengthening the depth, and being more adaptive to the chain evolution characteristics than traditional exponential decay.

[0051] After verification, if the score >0.7, mark the block as valid and broadcast the confirmation, otherwise trigger an alarm, record the violation log and isolate the related node, the whole process uses the Harmony soft bus to multicast the verification result and the chain copy, ensuring distributed consistency and minimal delay, e.g. <100ms, further strengthening the reliability of the audit through this verification mechanism, and providing a reliable basis for the subsequent destruction link.

[0052] S4: Based on the blockchain verification result, trigger the data destruction process through the smart contract, use the convolutional neural network for multiple compliance verification, calculate the destruction threshold, and execute the data cover removal when the conditions are met, generate the destruction proof and synchronize to the blockchain.

[0053] Specifically, first, based on the input blockchain verification result, activate the smart contract for destruction triggering. The specific operation is to parse the compliance score and block hash of the structured object, if the score <0.7, indicating non-compliance or end-of-life flag, call the smart contract deployed on the Harmony distributed soft bus, written in a language similar to Solidity, input the serialized block of the compliance instance as the event payload, the contract first deserializes the block inside, extracts the audit history such as permission changes and risk details, then verifies the block hash through ECDSA signature to match the local chain copy, ensuring no tampering. This activation process triggers through multi-node consensus execution of the contract, at least 3 nodes confirmed, avoiding single device failure, and gradually adjusts the trigger window according to the compliance score, e.g. 0.4-0.6 sets a 5-minute buffer period to allow chain synchronization, thus addressing the inconsistent destruction risk of distributed devices, such as health data remaining in edge nodes after expiration in wearable devices, leading to privacy regulation violations, improving the reliability of governance, reducing 35% asynchronous errors compared to general contract triggering, ensuring chain continuity through score-hash double verification, and improving the limitations of existing single-point destruction, improving distributed consistency.

[0054] Next, a multi-compliance verification is conducted to confirm the destruction is feasible, using a two-layer convolutional neural network, CNN, implemented on the Hyperion AI engine: the input layer receives the serialized block data of the compliance ledger as a one-dimensional sequence, with dimensions equal to the block size, e.g. 1024 bytes, converted to a floating-point tensor; the convolutional layer contains 16 filters with kernel size 3, using ReLU activation, scanning the chain in a mode like a sequence of consecutive block hashes to detect inconsistencies or residual violations; the pooling layer uses average pooling, outputting a verification vector with dimensions 4, representing the chain integrity score, the compliance consistency indicator, the risk residual probability, and the destruction readiness, respectively. The network is fine-tuned using local chain historical samples, embedding the node connectivity matrix of the soft bus query in the convolutional filters as dynamic kernel weights, to handle the distributed destruction peculiarities of Hyperion, such as partial loss of chain replicas across devices under network partition, leading to inaccurate pre-destruction verification, ensuring the verification more accurately reconstructs missing blocks, e.g. compensating for weak signal areas in a car moving scenario, thus providing a more robust final check, improving the static limitations of existing hash verification, and extending the filter range under low compliance scores, reducing the risk of residual data leakage, improving the overall privacy clearing efficiency, and better adapting to network partition scenarios than traditional verification methods.

[0055] Based on the verification vector and the blockchain verification result, the destruction threshold is calculated, using the following formula to decide the execution:

[0056] ;

[0057] where, is the destruction threshold, [0,1], > 0.8 triggers destruction; is the compliance score, directly taken from the blockchain verification result; is the verification vector average, computed from the CNN output, e.g. 0.6; is the block hash entropy, computed from the SHA-256 bit entropy of the compliance ledger, e.g. 0.8 indicates high randomness; is the chain length, from the model block number, e.g. 5; is a parameter. The formula incorporates the arctangent-quartic root growth term , the arctangent function normalizes the saturation effect of entropy, and combines with the fourth root gradual chain length, aiming at destroying the entropy-length coupling uncertainty on the distributed chain, such as the nonlinear growth risk of health data chain when expanding on multiple devices, providing flexible threshold adjustment, ensuring that destruction is more accurately prevented from premature removal, and slowly increasing under long chains to simulate the gradual constraints of the Mongolian resource allocation, improving privacy protection for automatic expiration destruction after long-term storage of medical data, reducing the number of variables while strengthening the depth, and being more suitable for chain expansion characteristics than traditional linear thresholds. Through this threshold mechanism, the limitations of existing destruction decisions are improved, and accuracy and efficiency are improved.

[0058] After the threshold is confirmed, the destruction is performed: the contract calls the Mongolian file system API to overwrite the relevant data storage, uses multiple rounds of zero padding erasure, such as 5 rounds of alternating writing of random bytes and zeros, to ensure that it is completely unrecoverable, and generates a destruction hash, SHA-256 of the state before and after erasure, as proof, the entire process is broadcasted and executed on the trigger node, and the result is appended to the tail of the chain as the final block, synchronized to all participating devices to ensure distributed consistent removal, and through this multiple round erasure and broadcast synchronization, the handling of residual risks is improved, and the recovery possibility is reduced compared to existing manual destruction methods, and the thoroughness of privacy protection is improved.

[0059] The embodiment of the present application also provides a data full life cycle security management device in a Mongolian system, comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the steps in the above-mentioned data full life cycle security management method embodiment in a Mongolian system, such as steps S1-S4 in the above-mentioned data full life cycle security management method embodiment in a Mongolian system. Figure 1 Or, the processor implements the functions of the modules in the above-mentioned system embodiments.

[0060] For example, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present application. The one or more modules can be a series of computer program instruction segments that can complete a specific function, which are used to describe the execution process of the computer program in the data full life cycle security management device in a Mongolian system.

[0061] The data full life cycle security management device in a Mongolian system can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The data full life cycle security management device in a Mongolian system can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the data full life cycle security management device in a Mongolian system can also include input and output devices, network access devices, buses, etc.

[0062] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of the data full life cycle security management device in the one Hongmeng system, and connects various parts of the data full life cycle security management device in the one Hongmeng system through various interfaces and lines.

[0063] The memory can be used to store the computer programs and / or modules, and the processor realizes various functions of the data full life cycle security management device in the one Hongmeng system by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application required by a function, etc.; and the data storage area can store data created according to the running of the air conditioner controller, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0064] If the module of the data full life cycle security governance equipment integration in the one-horizon system is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, an executable file, or some intermediate form. The computer-readable medium can include any entity or device that can carry the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0065] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned various method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0066] The above is the preferred embodiment of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered within the scope of protection of the present application.

Claims

1. A method for data life cycle security management in a Hongmeng system, characterized in that, The method comprises: standardizing preprocessing and feature extraction of input data, using a neural network model to extract data features, and classifying data based on an enhanced decision tree model and sensitivity calculation to generate classification labels and sensitivity scores; based on the classification results and distributed identity authentication, the automatic allocation of permissions is realized through the smart contract, and the preliminary permission policy is generated; based on the preliminary permission policy and classification label, device identity authentication and behavior analysis are performed, the access risk is dynamically evaluated using a recurrent neural network model, and the permission score is adjusted according to the access risk to generate a dynamic permission policy and an audit record; pack the audit record and dynamic permission policy as blockchain transaction data, store through a decentralized consensus protocol, and verify compliance based on a compliance calculation formula to generate a blockchain verification result and a compliance basis; based on the blockchain verification result, trigger the data destruction process through the smart contract, use convolutional neural network for multiple compliance verification, calculate the destruction threshold, and execute data cover cleaning when the conditions are met, generate destruction proof and synchronize to the blockchain.

2. The method according to claim 1, wherein the method is characterized in that, In the data classification, the sensitivity calculation includes a weighted combination of the base score, the distributed factor and the noise estimate, wherein the distributed factor is calculated based on the number of hops between devices.

3. The method of claim 1, wherein the method further comprises: In the automatic allocation of permissions, the calculation of the permission score includes a weighted combination of the sensitivity score, the identity matching degree and the device security, and a context correction term based on device load is introduced.

4. The method of claim 1, wherein the method further comprises: The recurrent neural network model is a four-layer LSTM structure; its input includes the frequency sequence of access requests, the timestamp sequence and the preliminary permission score; the network outputs a multi-dimensional behavior risk vector, which represents the frequency risk, the mode abnormal risk and the context interference risk.

5. The method of claim 1, wherein the method further comprises: The adjustment of the permission score is based on the preliminary permission score, the behavior risk factor, the identity verification score and the classification label factor, and is realized through a nonlinear function combination.

6. The method of claim 1, wherein the method further comprises: The audit record is generated in JSON format, including timestamp, visitor identity, permission change details, behavior risk data and classification label, and uses local encryption signature.

7. The method of claim 1, wherein the method further comprises: The storage process through the decentralized consensus protocol adopts the practical Byzantine fault tolerance protocol, and the number of fault-tolerant nodes is dynamically adjusted based on device load and audit record risk.

8. The method of claim 1, wherein the method further comprises: In the compliance verification, the calculation of the compliance score is based on the dynamic permission score, the average value of the behavior risk, the block height and the number of violation marks, and is evaluated through a nonlinear function.

9. The method of claim 1, wherein the method further comprises: receiving a request for a data object; determining whether the data object is stored in the data storage system; and if the data object is stored in the data storage system, retrieving the data object from the data storage system. In the data destruction process, the calculation of the destruction threshold is based on the compliance score, the average value of the verification vector, the block hash entropy and the chain length, and is adjusted through the inverse tangent function and the fourth root function.

10. The method of claim 1, wherein the method is implemented in a hyper- system. The data cleaning operation includes multiple rounds of cover writing with random bytes and zero values, and generates a destruction hash as an unforgeable destruction proof.

Images