Blockchain and federated learning based cross-institutional nursing data security sharing platform

Abstract

The present application relates to the technical field of health state monitoring and personalized intervention, in particular to a cross-institutional nursing data security sharing platform based on blockchain and federated learning, comprising: a parameter acquisition unit for monitoring and collecting the network average round-trip delay, effective network bandwidth and local data set size of each participant node, and for obtaining the average transaction confirmation time of the blockchain network; a time penalty quantization unit for calculating the effective synchronization time penalty index; an asynchronous anchoring decision unit for generating a regular anchoring signal; a periodic summary generation unit for aggregating key information in the anchoring period into a periodic summary for uploading to the blockchain; and an adaptive optimization unit for deciding the number of base rounds of the anchoring period for feedback adjustment; the present application discloses the inherent bucket effect of federated learning and the blockchain consensus overhead, converts the fuzzy system bottleneck problem into an accurate and quantifiable time cost index, and provides a scientific basis for subsequent dynamic regulation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health status monitoring and personalized intervention, in particular to a cross-institutional nursing data security sharing platform based on blockchain and federated learning. BACKGROUND

[0002] In the application scenario of fusing federated learning and blockchain technology to realize cross-institutional secure data sharing, it is crucial to ensure the auditability of the model training process; to achieve this goal, the computing process of federated learning and the consensus process of blockchain are usually synchronously coupled, that is, the key information of training is recorded on the chain in real time; however, in the actual distributed network, the network bandwidth, delay and local data size of each participating node are significantly heterogeneous, and the transaction confirmation time of the blockchain network is also in dynamic change; most existing technologies use fixed synchronization frequency or rely on manual configuration to coordinate off-chain computing and on-chain consensus, and this static coupling method lacks the ability to perceive environmental heterogeneity; when some nodes have poor network conditions or the on-chain transaction is congested, the strict synchronization mechanism will force the entire system to wait for the slowest link, causing the synchronization bottleneck to be dramatically amplified, severely restricting the overall training efficiency of federated learning; the existing method fails to establish a quantitative model that can comprehensively evaluate the time cost caused by multiple heterogeneous factors, so it cannot provide accurate basis for dynamically adjusting the synchronization strategy, making it difficult to achieve effective balance between training efficiency and process auditability.

[0003] Therefore, how to build a closed-loop adaptive regulation mechanism that can accurately quantify the system synchronization bottleneck and dynamically decouple the computing and consensus processes is a technical problem that needs to be solved in the field. SUMMARY

[0004] To solve the above technical problems, the present application discloses a cross-institutional nursing data security sharing platform based on blockchain and federated learning, specifically, the technical solution of the present application comprises:

[0005] A parameter acquisition unit is used to monitor and collect the network average round-trip delay, effective network bandwidth and local data set size of each participating node, and is used to obtain the average transaction confirmation time of the blockchain network;

[0006] A time penalty quantization unit is used to calculate the communication time cost of each participating node according to the network average round-trip delay, effective network bandwidth and local data set size collected by the parameter acquisition unit, and to calculate the effective synchronization time penalty index in combination with the average transaction confirmation time;

[0007] The asynchronous anchoring decision unit is configured to compare and analyze the effective synchronization time penalty index with a preset time penalty threshold; when the effective synchronization time penalty index is greater than the preset time penalty threshold, a frequency reduction anchoring signal is generated; and when the effective synchronization time penalty index is not greater than the preset time penalty threshold, a normal anchoring signal is generated.

[0008] The period summary generation unit is configured to perform off-chain model training in a corresponding anchoring period in response to the frequency reduction anchoring signal or the normal anchoring signal, and aggregate key information in the anchoring period into a period summary to upload to a blockchain.

[0009] The adaptive optimization unit is configured to monitor an actual total time consumption of the anchoring period, calculate an ideal time consumption of the anchoring period, and perform feedback adjustment on a basic round number used for decision of the anchoring period according to a ratio of the actual total time consumption to the ideal time consumption.

[0010] Preferably, the communication time cost is calculated as follows:

[0011] The local data set size is divided by the effective network bandwidth to obtain a quotient, and the quotient is summed with the network average round-trip delay to obtain a value, which is set as the communication time cost.

[0012] Preferably, the effective synchronization time penalty index is calculated as follows:

[0013] The maximum communication time cost among all participant nodes is obtained, and weighted processing is performed in combination with data heterogeneity; the processed value is summed with the weighted average transaction confirmation time to obtain a value, which is set as the effective synchronization time penalty index.

[0014] Preferably, the anchoring period is decided as follows:

[0015] The effective synchronization time penalty index is divided by the preset time penalty threshold, the obtained quotient is multiplied by the basic round number, and the product is rounded up to obtain a value, which is set as the anchoring period.

[0016] Preferably, the period summary generation unit is configured to concatenate a hash value of an aggregated model parameter at the end of a period, a hash value of a contribution proof of each participant in the period, and a start and end timestamp of the period, and perform hash function processing on the concatenated byte string to generate a period summary.

[0017] Preferably, the ideal time consumption is calculated as follows:

[0018] The number of rounds contained in the anchoring period of the current period is multiplied by the maximum communication time cost among all participant nodes, and the obtained product value is summed with the average transaction confirmation time to obtain a value, which is set as the ideal time consumption.

[0019] Preferably, the feedback adjustment process of the base round number is as follows:

[0020] The actual total time consumption is divided by the ideal time consumption, the obtained quotient is subjected to damping weighting processing, the processed result is multiplied by the base round number of the current period, and the obtained product is set as the base round number of the next period.

[0021] Preferably, the layered governance unit is further included;

[0022] The layered governance unit is used to obtain a preset proposal type influence factor, a reputation score of a proposal initiator and an estimated amount of global potential influence of the proposal, and calculate a governance importance score in combination with the proposal type influence factor, the reputation score of the proposal initiator and the estimated amount of global potential influence of the proposal.

[0023] Preferably, the layered governance unit is further used to trigger an on-chain voting process when the governance importance score exceeds a preset governance threshold, and not trigger the on-chain voting process when the governance importance score does not exceed the preset governance threshold.

[0024] Preferably, the reputation score of the proposal initiator is calculated in combination with the effectiveness of the contribution proof of the initiator in a historical anchor period and the node parameter stability reflected by the communication time cost.

[0025] The cross-institutional nursing data security sharing platform based on the blockchain and federated learning provided by the application aims to overcome the synchronization bottleneck amplification effect caused by the strong coupling of federated learning and the blockchain in a heterogeneous environment.

[0026] 1. The application establishes a mechanism for accurately quantifying system performance bottlenecks; the network delay, effective bandwidth, local data size and transaction confirmation time of the blockchain network of each participating node are monitored in real time by the parameter acquisition unit, and these heterogeneous parameters are uniformly modeled by the time penalty quantization unit to calculate an effective synchronization time penalty index that can comprehensively reflect the communication time cost and data heterogeneity influence; the index deeply reveals the inherent bucket effect of federated learning and the consensus overhead of the blockchain, and converts the vague system bottleneck problem into an accurate and quantifiable time cost indicator, providing a scientific basis for subsequent dynamic regulation, far exceeding the isolated and one-sided evaluation of performance bottlenecks in the prior art;

[0027] 2. The application constructs a flexible decoupling computing and consensus asynchronous anchoring decision and execution system; the asynchronous anchoring decision unit compares the aforementioned quantified time penalty index with the preset time penalty threshold to dynamically decide the anchoring signal that adapts to the current system state; the periodic summary generation unit responds to the signal to execute multiple rounds of off-chain model training within an anchoring period, and aggregates the periodic summary uploaded to the blockchain, including the aggregated model parameter hash, the proof of contribution hash of each party, and the timestamp; this design breaks the shackles of rigid synchronization with the blockchain in the traditional scheme, realizes the dynamic scaling of the on-chain evidence granularity, greatly reduces the interaction frequency with the blockchain, and significantly improves the overall efficiency of model training;

[0028] 3. The application introduces a closed-loop adaptive optimization feedback system; the adaptive optimization unit quantifies the performance deviation by monitoring the actual total time consumption of each anchoring period and comparing it with the ideal time consumption calculated based on the current system parameters; based on the deviation, the system can continuously feedback and adjust the basic round number that determines the length of the anchoring period; this mechanism gives the platform the ability to self-calibrate and evolve, enabling it to automatically adapt to the dynamic changes in network environment and data distribution, continuously maintaining the platform at the optimal balance point of model training efficiency and process auditability, solving the rigidity and lag problems caused by fixed parameters or relying on manual tuning in the prior art;

[0029] 4. The application designs a high-efficiency and secure hierarchical governance architecture; the hierarchical governance unit calculates the governance importance score by comprehensively evaluating the influence factors of the proposal type, the reputation score of the proposal initiator, and the potential impact of the proposal; this scoring system innovatively links the historical technical performance of the node, such as the effectiveness of the proof of contribution and the stability of communication, with its voice in community governance; based on this score, the system can automatically distribute different importance of proposals, trigger high-security on-chain voting for major impact matters, and handle routine transactions through efficient off-chain processes; this not only avoids congestion of all governance decisions in the inefficient on-chain voting channel, but also builds a positive incentive mechanism to encourage all participants to contribute stable and high-quality resources to the platform, thereby ensuring the long-term health and benign development of the entire data sharing ecosystem in a higher dimension. BRIEF DESCRIPTION OF DRAWINGS

[0030] The application will be further explained in conjunction with the drawings and examples:

[0031] Figure 1 is a structural diagram of the application. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below with specific examples.

[0033] Example 1:

[0034] Please refer to Figure 1 , a cross-institutional nursing data security sharing platform based on blockchain and federated learning, comprising:

[0035] A parameter collection unit is configured to monitor and collect the network average round-trip delay, effective network bandwidth and local data set size of each participant node, and is configured to obtain the average transaction confirmation time of the blockchain network;

[0036] A time penalty quantization unit is configured to calculate the communication time cost of each participant node according to the network average round-trip delay, effective network bandwidth and local data set size collected by the parameter collection unit, and to calculate the effective synchronization time penalty index in combination with the average transaction confirmation time;

[0037] An asynchronous anchoring decision unit is configured to compare and analyze the effective synchronization time penalty index with a preset time penalty threshold; when the effective synchronization time penalty index is greater than the preset time penalty threshold, a frequency reduction anchoring signal is generated; when the effective synchronization time penalty index is not greater than the preset time penalty threshold, a regular anchoring signal is generated;

[0038] A cycle summary generation unit is configured to perform off-chain model training in the corresponding anchoring period in response to the frequency reduction anchoring signal or the regular anchoring signal, and aggregate the key information in the anchoring period into a cycle summary for uploading to the blockchain;

[0039] An adaptive optimization unit is configured to monitor the actual total time consumption of the anchoring period, calculate the ideal time consumption of the anchoring period, and adjust the number of basic rounds used to decide the anchoring period according to the ratio of the actual total time consumption to the ideal time consumption;

[0040] The embodiment of the present application provides a cross-institutional nursing data security sharing platform based on blockchain and federated learning, comprising: a parameter collection unit, a time penalty quantization unit, an asynchronous anchoring decision unit, a cycle summary generation unit, and an adaptive optimization unit;

[0041] The embodiment aims to solve the technical pain point of the amplification effect of the synchronization bottleneck in the network and data heterogeneous environment due to the strong coupling of federated learning and blockchain synchronization in the prior art; the platform realizes the combination of model training efficiency and process auditability through a closed-loop adaptive system that dynamically decouples the calculation and consensus process;

[0042] Parameter collection unit, which aims to provide real-time, multi-dimensional data input for subsequent performance bottleneck quantification and dynamic strategy adjustment; in this embodiment, this unit is implemented through the deployment of agent programs on the data servers of each participating party client, such as each nursing institution; specifically, this unit is responsible for monitoring and collecting key parameters that affect synchronization performance, and its working mode is as follows:

[0043] Network average round-trip delay Effective network bandwidth : Obtained by periodically sending probe packets of a standard size from the client agent to the federated learning aggregator and recording the response time; Refers to the average time for a packet to go from the client to the aggregator and back, measured in seconds Its role is to measure the delay of network communication; Refers to the actual available data transfer rate within a certain time, measured in megabytes per second Its role is to measure the flow of network communication;

[0044] Size of local data set : The size of the local stored nursing data set file or the number of records is directly read by the local client agent, measured in megabytes Its role is to quantify the load of data transmission;

[0045] Average transaction confirmation time : Obtained by actively querying the API interface of the blockchain network accessed, reflecting the congestion and consensus efficiency of the current blockchain network, measured in seconds Its role is to quantify the time cost of on-chain storage;

[0046] Time penalty quantification unit, which aims to unify multiple heterogeneous performance influencing factors into a mathematical model with clear physical meaning and predictable time cost index; in this embodiment, this unit constructs a mathematical model called effective synchronization time penalty index based on the data collected by the parameter collection unit; to clarify its internal logic, this unit first calculates the communication time cost of each participating party node n Its calculation method is as follows:

[0047]

[0048] Where, Represents the estimated total communication time required by node n to complete one federated learning data interaction, such as uploading model updates, measured in seconds Its source is the previous , , The parameter is calculated, and the formula combines network delay and data transmission time, fully reflecting the communication performance of the node.

[0049] Based on the above communication time cost, the unit combines the worst case among all participants and the overhead of the blockchain to calculate the effective synchronization time penalty index Ψ:

[0050]

[0051] Ψ represents the expected total time cost of the system to complete a round of fully synchronized federated learning interaction, including off-chain computing and on-chain consensus, in seconds , which is calculated by the unit; It embodies the wooden bucket effect that federated learning must wait for the slowest node; Data heterogeneity is a dimensionless pure number obtained by calculating the KL divergence of local data distribution and global data distribution, and its source is the historical value evaluated by the federated learning framework after the training of the last anchor period. Its role is to quantify the impact of data distribution difference on model convergence speed; It is a dimensionless weight coefficient, and its calibration process is crucial to ensure the accuracy of the model. To clarify its calibration method, it is necessary to distinguish between the variables used in the calibration process and the model runtime. Calibration is performed in a benchmark test environment, which includes a set of controllable test nodes. In the calibration phase, set multiple different network parameter combinations and data set sizes for each test node , and record the actual total time to complete a round of synchronization interaction ; Thus, a set of data points consisting of independent variables and dependent variables can be obtained, representing the average network round-trip delay in the benchmark test environment, representing the effective network bandwidth in the benchmark test environment, representing the local data set size in the benchmark test environment; By performing multiple linear regression fitting on these data points, the best estimate values of and are obtained; This index converts the vague performance bottleneck problem into a quantifiable time cost, providing an accurate basis for subsequent asynchronous decision-making;

[0052] The asynchronous anchor decision unit aims to dynamically adjust the granularity of blockchain storage based on the current estimated time cost of the system, breaking the rigid synchronization relationship between computing and consensus. In this embodiment, the unit compares and analyzes the effective synchronization time penalty index Ψ calculated by the time penalty quantification unit with a preset time penalty threshold ; The maximum single round synchronization time cost acceptable by the system, for example, 10 seconds, which serves as the basis for triggering the asynchronous strategy, and is determined by experience based on the real-time requirements of the business and the stress test of the system performance;

[0053] When , it means that the current system will produce unacceptable time delay if it adopts the synchronous mode, and the unit generates the frequency reduction anchor signal at this time;

[0054] When , it means that the system performance is still within the acceptable range, and the unit generates the regular anchor signal at this time;

[0055] Period summary generation unit, which is used to perform specific off-chain calculation and to trustfully solidify the key process information in a lightweight manner within a calculation period; in this embodiment, the unit performs model training and aggregation of N rounds of federated learning within a time window called anchor period in response to the signal generated by the asynchronous anchoring decision unit, and this process is completely completed off-chain and is not affected by the delay of the blockchain consensus; the anchor period refers to a group of consecutive federated learning communication rounds, and after the end of the anchor period, an interaction with the blockchain is performed, which bundles multiple calculations and amortizes the cost of single on-chain; at the end of the period, the unit aggregates all the key information in the period, such as the global model parameter hash , the hash of the contribution proof of each participant, and the start and end time stamps of the period , into a single period summary through an anti-collision hash function such as SHA-256, and uploads the summary as transaction content to the blockchain, while the detailed log is saved in the off-chain distributed storage;

[0056] Adaptive optimization unit, which is used to build a top-level closed-loop feedback mechanism to enable the system to self-regulate the core control parameters according to the actual running performance, so as to dynamically optimize between efficiency and auditability; in this embodiment, after the end of each anchor period, the unit first monitors and records the actual total time consumption of the period ; at the same time, it calculates the ideal time consumption of an anchor period based on the actual communication time cost of each node in the period and the blockchain confirmation time as the benchmark for performance evaluation; finally, the unit adjusts the number of basic rounds for decision of the next anchor period according to the ratio of the actual total time consumption to the ideal time consumption ;

[0057] The application builds a complete closed-loop adaptive system through the cooperation of the above units; it first accurately estimates the cost of the synchronization bottleneck by parameter acquisition and time penalty quantization, then dynamically decides the decoupling degree of calculation and consensus based on the cost, and greatly reduces the interaction frequency on the chain under the premise of ensuring traceability through periodic summary; most importantly, the adaptive optimization unit enables the system to learn from historical performance and continuously calibrate core parameters , so as to maintain the best balance point of efficient model training and reliable process record in complex and variable network and data environment, and solve the synchronization bottleneck amplification effect problem of the prior art.

[0058] Embodiment 2:

[0059] The calculation process of the communication time cost is as follows:

[0060] Divide the size of the local data set by the effective network bandwidth to obtain a quotient, and sum the quotient and the network average round-trip delay to obtain a value, which is set as the communication time cost;

[0061] On the basis of the platform described in Embodiment 1, the calculation process of the communication time cost is as follows: divide the size of the local data set by the effective network bandwidth to obtain a quotient, and sum the quotient and the network average round-trip delay to obtain a value, which is set as the communication time cost;

[0062] This embodiment is a specific implementation of the calculation method of the communication time cost in the time penalty quantization unit; as described above, the purpose is to accurately estimate the total time required for a single node to complete one data interaction, and the specific calculation formula is:

[0063]

[0064] The calculation logic of this formula is derived from the physical decomposition of the network communication process; the total time of one data interaction is mainly composed of two parts: one is the delay time of data packet transmission in the network, i.e. the network average round-trip delay ; the other is the transmission time required for the transmission of data content itself; the transmission time is determined by the size of the local data set (total data amount) and the effective network bandwidth (transmission rate), and the transmission time is obtained by dividing the two ; by adding the two independent time overheads, a comprehensive communication time cost with a unit of seconds is obtained;

[0065] wherein all parameters , , All are acquired in real time by the parameter acquisition unit through the aforementioned probe measurement and local reading mode, ensuring that the calculation result can truly reflect the current actual state of the node;

[0066] Through this calculation method combining delay and transmission bandwidth impact, the application can more accurately depict the communication performance of the node; compared with the scheme considering only delay or bandwidth single index, the scheme is more comprehensive and accurate in evaluating the performance of the node, for example, it can distinguish between two different performance bottleneck types of high network delay but large bandwidth and low network delay but small bandwidth, thereby providing a more reliable data basis for accurate construction of the upper layer effective synchronization time penalty index Ψ, and further improving the accuracy of dynamic decision-making of the whole system.

[0067] It should be noted that this formula is a first-order approximation model of communication time cost, which effectively captures the two core influencing factors of network delay and transmission flux. In actual application, time costs such as local calculation overhead, data serialization, etc. can be considered as contained in the network average round-trip delay This model provides accurate cost estimation while ensuring simple calculation, to support upper-level decision-making.

[0068] Embodiment 3:

[0069] The solving process of the effective synchronization time penalty index is as follows:

[0070] Obtain the maximum communication time cost in all participant nodes, and perform weighted processing in combination with data heterogeneity; sum the processed value and the weighted average transaction confirmation time, and set the obtained value as the effective synchronization time penalty index;

[0071] On the basis of the platform described in Embodiment 1, the solving process of the effective synchronization time penalty index is as follows: obtain the maximum communication time cost in all participant nodes, and perform weighted processing in combination with data heterogeneity; sum the processed value and the weighted average transaction confirmation time, and set the obtained value as the effective synchronization time penalty index;

[0072] This embodiment is a specific implementation of the solving process of the effective synchronization time penalty index Ψ in the time penalty quantization unit; as described above, the purpose of Ψ is to estimate the total time cost required by the system to complete a whole round of synchronization interaction, and its specific solving formula is:

[0073]

[0074] The internal logic of this formula integrates the bucket effect of federated learning, the influence of data heterogeneity, and the independent overhead of blockchain consensus;

[0075] Quantifying the "barrel effect": In federated learning, each round of aggregation must wait for the slowest participant to complete its model upload. Therefore, the solution process first... The operation is to obtain the largest communication time cost among all participating nodes. This serves as the benchmark for communication time in this round of the chain;

[0076] Impact of data heterogeneity: Data heterogeneity This refers to the degree of inconsistency in the data distribution among the participating parties. Specifically, if we assume that the participating parties... The local data distribution is The global data distribution is as follows Then data heterogeneity can be defined as ,in, This represents the Kullback-Leibler divergence, used to measure the difference between the local and global data distributions. At the end of each anchoring cycle, the system estimates the current global distribution based on the model updates uploaded by each party. and the local distribution known to each participant. Compare, and thus update Value. During the initial anchoring period of the system, It can be initialized to 0, indicating that no prior assumptions are made about the data distribution. If, for some reason, a valid result cannot be successfully calculated after a certain period ends... The system can either reuse the historical value from the previous cycle or reset it to 0 to ensure the robustness of the decision-making process.

[0077] It increases the number of training epochs required for the model to converge, which is simplified here as a weighted effect on the time cost of a single epoch; by multiplying by This term quantifies the negative impact of data distribution differences on training efficiency as an increase in time cost; to ensure the uniqueness of the computation, this... Specifically refers to The data heterogeneity index of the node with the highest determined communication time cost.

[0078] The linear weighting here is an engineering simplification that maps the convergence problem at the algorithm level to the time cost at the system level. Its purpose is not to accurately predict the total number of rounds required for convergence, but rather to assign higher risk weights to nodes with poorer data distribution at the granularity of a single round of decision-making, thereby reflecting the additional training overhead that may result from data heterogeneity in the time penalty exponent. This simplification is effective for macro-level control.

[0079] Multi-source cost fusion: This involves weighting communication penalties based on data heterogeneity. With weight Adjusted average transaction confirmation time The two time overheads are serial, so the total estimated time cost Ψ can be obtained by adding them up;

[0080] The solution process is not simply adding up individual time parameters, but rather it deeply understands and models the core bottleneck sources in distributed collaborative systems; by adopting Instead of the average value, the effect of the wooden barrel is accurately captured; by introducing The data heterogeneity problem at the algorithm level and the time cost problem at the system level are directly mathematically related; this makes the Ψ index not only a time estimate, but also a multi-dimensional comprehensive evaluation of the overall health of the system, which is much more accurate than existing technologies that look at network latency or transaction time in isolation, providing extremely accurate decision input for subsequent asynchronous anchoring strategies.

[0081] Embodiment 4:

[0082] The decision-making process of the anchoring period is as follows:

[0083] Divide the effective synchronization time penalty index by the preset time penalty threshold, multiply the quotient by the number of base rounds, and round up the product to obtain the value of the anchoring period;

[0084] On the basis of the platform described in Embodiment 1, the decision-making process of the anchoring period is as follows: divide the effective synchronization time penalty index by the preset time penalty threshold, multiply the quotient by the number of base rounds, and round up the product to obtain the value of the anchoring period;

[0085] This embodiment is a specific embodiment of the decision-making process of how the asynchronous anchoring decision-making unit determines the anchoring period N, and the specific decision-making formula is:

[0086]

[0087] The decision-making logic is derived from the threshold triggering and proportional regulation mechanism in control theory, aiming to dynamically adjust the interaction frequency with the blockchain;

[0088] By the ratio , the current estimated effective synchronization time penalty index Ψ is evaluated relative to the system's acceptable time penalty threshold ; if , the ratio is not greater than 1; if , the ratio is greater than 1, and the larger the value represents the more serious the performance bottleneck;

[0089] Multiply this ratio by a base round number ; the base round number refers to the ideal state of the system, that is, Ψ is approximately equal to The reference anchor frequency, for example, can be set to 5 rounds; its initial value The basic requirement of the audit granularity according to the business and the performance evaluation result in the benchmark test environment are comprehensively determined as the starting point of the system adaptive adjustment; when the system performance is good, the anchor period N is taken as ; when the system performance bottleneck occurs, the anchor period N will increase linearly with the increase of Ψ;

[0090] The ceiling function is used to ensure that the calculated number of rounds N is a positive integer that can be executed;

[0091] This decision-making process establishes a direct and automatic control channel from performance bottleneck quantification to system behavior adjustment; when the system time cost Ψ rises, the anchor period N automatically becomes longer, thereby intelligently reducing the interaction frequency with the blockchain, effectively relieving the performance pressure; conversely, when the performance recovers, N will automatically shorten to pursue a more fine-grained audit granularity; this dynamic and quantitative decision-making mechanism enables the system to be flexible in strong and weak situations, completely avoiding the rigidity and lag problems caused by fixed anchor periods or manual parameter adjustment in the prior art.

[0092] Embodiment 5:

[0093] The period summary generation unit is used to concatenate the hash value of the aggregated model parameters at the end of the period, the hash value of the contribution proof of each participant in the period, and the start and end timestamps of the period, and perform hash function processing on the concatenated byte string to generate a period summary;

[0094] On the basis of the platform described in Embodiment 1, the period summary generation unit is used to concatenate the hash value of the aggregated model parameters at the end of the period, the hash value of the contribution proof of each participant in the period, and the start and end timestamps of the period, and perform hash function processing on the concatenated byte string to generate a period summary;

[0095] This embodiment is a specific implementation of how the period summary generation unit generates a period summary ; the process aims to compress all key information in an anchor period into a fixed-length, tamper-proof lightweight data to upload to the blockchain, and the specific generation formula is:

[0096]

[0097] This generation logic draws on the Merkle tree idea in cryptography to ensure data integrity and verifiability;

[0098] wherein, : This is a standard anti-collision hash function, such as SHA-256, which can convert an input of arbitrary length into a fixed-length unique digest;

[0099] : : This refers to the hash value of the parameter file of the global model generated by the final aggregation at the end of the anchoring period; its role is to solidify the final calculation results of this period;

[0100] : : This refers to the hash value of the local model update or other contribution proof of each round of contribution by the participant m in this period; first, the contribution proof of each participant is hashed, and then it is included in the overall digest; its role is to ensure that the contribution of each participant can be traced and cannot be denied;

[0101] : : This refers to the start and end timestamps of the anchoring period; its role is to provide a clear temporal context for the digest;

[0102] : This symbol represents the sequential concatenation operation of byte strings; the unit concatenates all the above data items in a predetermined order into a long byte string;

[0103] : Perform a hash function on the entire long byte string : The final result is the period digest ;

[0104] : The period digest generated in this way has a very high information compression rate and strong security features; it only occupies a small space of a blockchain transaction, which is equivalent to solidifying all the key information of N rounds of training in the period; any tampering with the detailed logs saved off-chain, such as a certain round of model update by a participant, will result in a period digest : The value recorded on the chain is inconsistent, which can be easily discovered; this mode of on-chain storage and off-chain detailed description reduces the amount of data uploaded to the chain by N orders of magnitude without sacrificing the final auditability, and is one of the core technical means to decouple computation and consensus and greatly improve the efficiency of system operation.

[0105] Embodiment 6:

[0106] : The ideal time-consuming calculation process is as follows:

[0107] : Multiply the number of rounds included in the anchoring period of the current period by the maximum communication time cost among all participant nodes, and sum the product value with the average transaction confirmation time to obtain the value, which is set as the ideal time consumption;

[0108] On the basis of the platform described in Embodiment 1, the calculation process of the ideal time consumption is as follows: multiply the number of rounds contained in the anchor period of the current period by the maximum communication time cost in all participant nodes, sum the product value with the average transaction confirmation time, and set the obtained value as the ideal time consumption;

[0109] The purpose of this embodiment is to calculate the ideal time consumption of the adaptive optimization unit The specific calculation process of the ideal time consumption is as follows: The purpose of this embodiment is to provide a reasonable performance benchmark based on the current system state for the actual performance of the current kth anchor period, and the specific calculation formula is as follows:

[0110]

[0111] The formula aims to simulate the time required to complete the current anchor period in an ideal case;

[0112] By Obtaining the maximum communication time cost in all participant nodes; this is consistent with the system's understanding of the bottleneck effect, representing the expected time in the real case that must wait for the slowest node to complete a round of synchronization;

[0113] Multiply the maximum time consumption by the actual number of rounds in the current kth period ; The actual number of rounds executed in the current period, whose value is determined by the decision-making process of Embodiment 4 at the end of the previous period; the product represents the total time required to complete all off-chain calculations of the current period in theory;

[0114] Add the average transaction confirmation time required for one on-chain operation ; since an anchor period only on-chains once, only add once

[0115] The key advantage of this ideal time consumption calculation method is its self-consistency and completeness; it does not use any fixed, preset ideal time parameters, but dynamically calculates based on the and values measured in real time by the parameter acquisition unit, reflecting the current real system state; this means that the performance benchmark will automatically adjust with changes in network conditions and blockchain congestion, allowing for a more fair and accurate evaluation of the actual total time consumption ; The establishment of this dynamic benchmark provides more reliable input for subsequent feedback adjustment of the basic number of rounds, avoiding incorrect adjustments due to rigid benchmarks.

[0116] Embodiment 7:

[0117] The feedback adjustment process for the number of base rounds is as follows:

[0118] Divide the actual total time by the ideal time, apply damped weighting to the quotient, multiply the result by the number of base rounds in the current cycle, and set the product as the number of base rounds in the next cycle.

[0119] Based on the platform described in Example 1, the feedback adjustment process for the number of base rounds is as follows: divide the actual total time by the ideal time, perform damped weighting on the quotient, multiply the result by the number of base rounds in the current cycle, and set the product as the number of base rounds in the next cycle.

[0120] This embodiment describes the number of rounds for the core control parameters in the adaptive optimization unit. The feedback adjustment process is concretized; this process is the top-level optimization logic of the system, constituting a closed-loop adaptive adjustment between efficiency and auditability, and its specific adjustment formula is as follows:

[0121]

[0122] This logic originates from the proportional feedback control concept in control theory and is implemented through a multiplicative form, making it more robust in a distributed environment.

[0123] Through ratio Calculate the performance deviation for this cycle; This is the actual time spent. The ratio represents the ideal time taken; if the ratio is greater than 1, it indicates that the system is running slower than expected; if the ratio is less than 1, it indicates that there is performance margin.

[0124] Perform this performance deviation ratio Damped weighted processing of powers of 1; It is a dimensionless damping coefficient, typically ranging from 0 to 1, such as 0.5. Its function is to prevent the system from oscillating due to overexcited response. Its value is determined through system simulation experiments, inputting a series of typical step or impulse performance disturbances and observing different... Values The convergence rate and overshoot of the parameters are ultimately selected to enable the system to reach stability as quickly as possible while minimizing oscillations. value;

[0125] Multiply the performance deviation after damping weighting by the number of base rounds in the current cycle. The result obtained This is the new base round number set for the next cycle decision;

[0126] when If the ratio is greater than 1, the system will automatically increase it. This means that in the future, under similar Ψ index, the calculated anchoring period N will be longer, thus sharing the time cost by reducing the uplink frequency; conversely, when , the system will appropriately lower , and pursue a more refined anchoring period when performance is abundant;

[0127] The establishment of this feedback adjustment mechanism gives the system the ability to evolve itself; it no longer relies on a set of static, artificially adjusted parameters, but can automatically and continuously optimize its core asynchronous anchoring strategy according to real-world performance feedback; this closed-loop adaptive ability ensures that the system can long-term, stable operation at the best dynamic balance point of efficiency and credibility under any external environment, which is a major technological progress that existing static or open-loop systems cannot match;

[0128] To ensure the robustness of the algorithm, in actual calculation, a small positive lower limit is set for the ideal time cost , for example second, that is, the actual denominator used is This can prevent the calculation overflow problem caused by the denominator approaching zero in rare extreme cases, ensuring that the adjustment mechanism behaves as expected under all possible inputs.

[0129] Embodiment 8:

[0130] The platform also includes a hierarchical governance unit;

[0131] The hierarchical governance unit is used to obtain the preset proposal type influence factor, the reputation score of the proposal initiator, and the estimated amount of global potential impact of the proposal; and combined with the proposal type influence factor, the reputation score of the proposal initiator, and the estimated amount of global potential impact of the proposal, the governance importance score is calculated;

[0132] The hierarchical governance unit is also used to trigger the on-chain voting process when the governance importance score exceeds the preset governance threshold; and not trigger the on-chain voting process when the governance importance score does not exceed the preset governance threshold;

[0133] The proposal initiator's reputation score is calculated based on the effectiveness of the initiator's contribution proof in the historical anchoring period, and the stability of the node parameters reflected by the communication time cost;

[0134] On the basis of the platform described in Embodiment 1, a hierarchical governance unit is further included; the hierarchical governance unit is configured to obtain a preset proposal type influence factor, a reputation score of a proposal initiator, and an estimated amount of potential global influence of the proposal; and the governance importance score is calculated in combination with the proposal type influence factor, the reputation score of the proposal initiator, and the estimated amount of potential global influence of the proposal; the hierarchical governance unit is further configured to trigger an on-chain voting process when the governance importance score exceeds a preset governance threshold; the reputation score of the proposal initiator is calculated based on the effectiveness of the contribution proof of the initiator in the historical anchoring period and the stability of the node parameter reflected by the communication time cost.

[0135] The embodiment is a specific implementation of the hierarchical governance unit introduced by the platform at the management level; the purpose of the unit is to distinguish between high-frequency operational transactions and low-frequency but significant governance decisions, and to design an efficient and secure automated decision flow mechanism for the latter;

[0136] The core of the unit is to calculate a governance importance score The calculation formula is derived from the multi-criteria decision analysis model:

[0137]

[0138] Among them, : the governance importance score, a normalized dimensionless value, which is calculated by the unit;

[0139] : the proposal type influence factor, the preset influence weight according to the type of the governance proposal, such as the addition of a new institution, the release of a global model version, the modification of system parameters, etc., which is a constant set by the alliance governance committee according to the governance rules during system initialization;

[0140] : the reputation score of the proposal initiator, which is the key to the parameter linkage design between modules; the reputation score is not a static value, but a dynamically calculated value; the calculation basis includes: a) the effectiveness of the contribution proof submitted by the initiator in the historical anchoring period, which is judged by whether the is passed in the subsequent verification generated by the cycle digest generation unit; b) the stability of the node of the initiator, which is reflected by the fluctuation of the historical communication time cost of the node; a node with a real historical contribution and a stable network will have a higher reputation score;

[0141] : the estimated amount of potential global influence of the proposal, which is quantitatively evaluated according to the template by the proposal initiator when submitting the proposal, and is preliminarily reviewed by humans or automatically.

[0142] : corresponding dimensionless weight coefficient, set by the alliance governance committee, for balancing the importance of different factors;

[0143] The score is calculated After scoring, the hierarchical governance unit compares it with a preset governance threshold The threshold is set according to the governance risk control strategy, for example, the governance committee can set a target to require the top 10% of proposals initiated by low-reputation nodes or with the largest estimated influence to be forced into on-chain voting, and then calculate the specific value of by statistical analysis of historical proposal data;

[0144] When , it indicates that the proposal is important, and the unit automatically triggers the on-chain smart contract to start a formal voting process that requires multi-signature confirmation;

[0145] When , it indicates that the proposal has a small impact or is a routine operation, which is handled by an off-chain lightweight consensus mechanism or directly approved by a specific role administrator, without the need for high-cost on-chain voting;

[0146] The introduction of the hierarchical governance model greatly improves the efficiency and security of the system in the management layer; it separates transactions of different properties to the most efficient processing path through the scoring mechanism, avoiding the congestion of all decisions in the inefficient on-chain voting channel; the most innovative point is to directly link the historical technical performance of the node to its governance authority; this design not only makes the governance decision more fair and credible, but also positively encourages all participants to actively and stably contribute high-quality data and computing resources to federated learning from the economic and game theory perspective, thereby ensuring the long-term healthy development of the entire platform and the virtuous cycle of the data sharing ecosystem in a higher dimension.

[0147] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A blockchain and federated learning based cross-institutional nursing data security sharing platform, characterized in that, The application comprises the following steps: A parameter collection unit is configured to monitor and collect the network average round-trip delay, effective network bandwidth and local dataset size of each participant node, and obtain the average transaction confirmation time of the blockchain network; A time penalty quantification unit is configured to calculate the communication time cost of each participant node according to the network average round-trip delay, effective network bandwidth and local dataset size collected by the parameter collection unit, and solve the effective synchronization time penalty index in combination with the average transaction confirmation time; An asynchronous anchoring decision unit is configured to compare and analyze the effective synchronization time penalty index with a preset time penalty threshold; when the effective synchronization time penalty index is greater than the preset time penalty threshold, a frequency reduction anchoring signal is generated; when the effective synchronization time penalty index is not greater than the preset time penalty threshold, a regular anchoring signal is generated; A periodic summary generation unit is configured to perform off-chain model training in the corresponding anchoring period in response to the frequency reduction anchoring signal or the regular anchoring signal, and aggregate the key information in the anchoring period into a periodic summary to upload to the blockchain; An adaptive optimization unit is configured to monitor the actual total time consumption of the anchoring period, calculate the ideal time consumption of the anchoring period, and feedback adjust the basic round number used to decide the anchoring period according to the ratio of the actual total time consumption to the ideal time consumption; The calculation process of the communication time cost is as follows: The local dataset size is divided by the effective network bandwidth to obtain a quotient, and the quotient is summed with the network average round-trip delay to obtain the communication time cost; The solving process of the effective synchronization time penalty index is as follows: The maximum communication time cost among all participant nodes is obtained and weighted according to data heterogeneity; the processed value is summed with the weighted average transaction confirmation time to obtain the effective synchronization time penalty index; The decision process of the anchoring period is as follows: The effective synchronization time penalty index is divided by the preset time penalty threshold, the quotient is multiplied by the basic round number, and the product is rounded up to obtain the anchoring period. 2.The blockchain and federated learning based cross-institutional care data security sharing platform of claim 1, wherein, The periodic summary generation unit is configured to concatenate the hash value of the aggregated model parameter at the end of the period, the hash value of the contribution proof of each participant in the period, and the start and end time stamp of the period, and perform hash function processing on the concatenated byte string to generate a periodic summary. 3.The blockchain and federated learning based cross-institutional care data security sharing platform of claim 1, wherein, The calculation process of the ideal time consumption is as follows: The number of rounds contained in the anchoring period of the current period is multiplied by the maximum communication time cost among all participant nodes, and the product is summed with the average transaction confirmation time to obtain the ideal time consumption. 4.The blockchain and federated learning based cross-institutional care data security sharing platform of claim 1, wherein, The feedback adjustment process of the basic round number is as follows: The actual total time consumption is divided by the ideal time consumption, the quotient is damped and weighted, the processed result is multiplied by the basic round number of the current period, and the product is set as the basic round number of the next period. 5.The blockchain and federated learning based cross-institutional care data security sharing platform of claim 1, wherein, The application further comprises a hierarchical governance unit. The hierarchical governance unit is configured to obtain a preset proposal type influence factor, a reputation score of a proposal initiator, and an estimated amount of potential global influence of the proposal; and calculate a governance importance score based on the proposal type influence factor, the reputation score of the proposal initiator, and the estimated amount of potential global influence of the proposal. 6.The blockchain and federated learning based cross-institutional care data security sharing platform of claim 5, wherein, The hierarchical governance unit is further configured to trigger an on-chain voting process when the governance importance score exceeds a preset governance threshold, and not trigger the on-chain voting process when the governance importance score does not exceed the preset governance threshold.

7. The blockchain and federated learning based cross-institutional care data secure sharing platform according to claim 5, characterized in that, The reputation score of the proposal initiator is calculated based on the validity of the contribution proof of the initiator in a historical anchoring period and the stability of a node parameter reflected by a communication time cost.

Images