Student growth trajectory evidence storage and tracing system based on hash chain tamper-proofing

Abstract

The present application relates to the field of education informatization and data evidence technology, in particular to a student growth track evidence and tracing system based on hash chain tamper-proofing, which comprises business data access, double time label standardization, correlation determination, delay buffer, batch output, hash graph construction, Merkle tree evidence and query tracing; after accessing the original event, the event occurrence time, system receiving time and delay time are generated, the main batch, backfilling or side branch are determined by combining the event correlation degree and sliding window, the current period directed acyclic graph and cycle root are formed and the main chain is anchored, and the growth track is reconstructed after checking the original content hash, double time label and source identification according to the student identification during query; the present application is used for reducing the event logic causal misplacement rate caused by disorderly access.

Description

Technical Field

[0001] This invention relates to the field of educational informatization and data storage technology, specifically a student growth trajectory storage and traceability system based on hash chain anti-tampering. Background Technology

[0002] Existing university data management systems typically need to aggregate multi-source data from academic affairs, student affairs, evaluation, practical activities, and IoT devices, and then unify, organize, and store this data.

[0003] In related technologies, multi-source data suffers from reporting delays and out-of-order arrival during transmission. Relying solely on the system's receiving order to solidify data can lead to distortion of the logical sequence of events, thereby affecting the consistency of growth trajectory reconstruction, correlation judgment, and subsequent evidence preservation.

[0004] Therefore, the multi-source event processing methods in related technologies cannot simultaneously achieve real-time access, real-order reconstruction, and reliable evidence preservation. Summary of the Invention

[0005] The purpose of this invention is to provide a student growth trajectory storage and traceability system based on hash chain anti-tampering, which avoids the distortion of the relationship between events caused by processing only according to the order of receipt, and is easier to take into account the real-time access of multi-source data, reconstruction of the true order and reliable storage.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] The student growth trajectory storage and traceability system based on hash chain anti-tampering includes the following modules connected in sequence: business data access module, dual time tag standardization module, association judgment module, delay buffer module, hash graph construction module, Merkle tree storage module, and on-chain anchoring and traceability verification module.

[0008] The business data access module is used to collect raw events, which include the event occurrence time, system reception time, and event payload data.

[0009] The dual-time-tag standardization module is used to determine the late time based on the event occurrence time and system reception time of the original event, and generate event records with dual timestamps;

[0010] The correlation determination module is used to determine the correlation degree of events based on event payload data, and to generate event handling decisions based on the lateness time and event correlation degree.

[0011] The delay buffer module is used to buffer event records within a sliding window. When the window expires, the current batch meets the preset closing conditions, or the capacity limit is reached, the batch is sealed and the set of nodes to be solidified is output.

[0012] The hash graph construction module is used to sort the set of nodes to be solidified by event occurrence time and event processing decision, generate node hashes for events that match the parent node set according to preset dependencies, construct the directed acyclic graph of the current period, and output the hashes of closed nodes.

[0013] The Merkle tree notarization module is used to generate a periodic root and a hash proof certificate based on the hash of the closed node; the on-chain anchoring and traceability verification module is used to append the periodic root to the main anchoring chain record, and output the student's growth trajectory based on the main anchoring chain, the periodic root, the current periodic directed acyclic graph, and the hash proof certificate when a traceability request is received.

[0014] In one alternative: the association determination module includes a main judgment unit, a backfilling judgment unit, a side branch judgment unit, and an anomaly alarm unit;

[0015] The main batch decision unit is used to generate the main batch entry decision when the event correlation is greater than or equal to the main chain entry threshold and the late time is less than or equal to the sliding window threshold.

[0016] The backfill determination unit is used to generate a backfill decision when the event correlation is greater than or equal to the main chain batch threshold and the sliding window threshold is less than the late time, and the late time is less than or equal to the product of the late amplification coefficient and the sliding window threshold; or when the side branch retention threshold is less than or equal to the event correlation and the event correlation is less than the main chain batch threshold, and the late time is less than or equal to the product of the late amplification coefficient and the sliding window threshold.

[0017] The side branch determination unit is used to generate a side branch diversion decision when the late arrival time is greater than the product of the late arrival amplification factor and the sliding window threshold, or when the event correlation is less than the side branch retention threshold.

[0018] The exception alarm unit is used to generate a discard alarm decision when the field integrity and format validation of the event record fails.

[0019] The association determination module outputs the main batch entry decision, the replenishment entry decision, the side branch diversion decision, or the discard alarm decision as event processing decisions. Among them, the main batch entry threshold, the side branch retention threshold, the late arrival amplification coefficient, and the sliding window threshold are all preset configuration parameters, and the main batch entry threshold is greater than the side branch retention threshold, and the late arrival amplification coefficient is greater than 1.

[0020] In one alternative: the delay buffer module includes a candidate tagging unit, a batch triggering unit, and a set output unit;

[0021] The candidate tagging unit is used to receive event records with dual time tags, and based on the event processing decision, it tags the event records as main batch candidate records, supplementary candidate records or side branch candidate records, and stores them in the buffer pool.

[0022] The batch sealing trigger unit is used to trigger a batch sealing command when the waiting time in the buffer pool reaches the sliding window threshold, or the event records in the current batch meet the preset closing conditions, or the number of event records in the batch reaches the capacity limit.

[0023] The set output unit is used to output the event records that meet the closure conditions as a set of nodes to be solidified after receiving the sealing and batching instruction, and to retain the event records that do not meet the closure conditions to the next round of buffer pool.

[0024] In one alternative: the hash graph building module includes: sorting events in the set of nodes to be solidified according to the event occurrence time and event processing decision by a sorting unit;

[0025] The parent node matching unit filters the set of parent nodes for each sorted event record according to a preset dependency relationship;

[0026] When not all parent nodes are matched and the event's lateness is less than or equal to the preset recovery time window, the corresponding event record is temporarily stored as a node to be attached by attaching the temporary storage unit.

[0027] The node hash is generated by the node hash generation unit when the parent nodes are complete, and written into the directed acyclic graph of the current period.

[0028] When the delay time of the event record exceeds the allowable time window for backfilling and still cannot match all parent nodes for closure, the side branch transfer unit transfers the corresponding node into the side branch directed acyclic graph.

[0029] In one alternative: the Merkle tree notarization module includes a dependency verification unit, a content verification unit, a temporary root generation unit, and a formal root upgrade unit;

[0030] The dependency verification unit is used to verify whether all nodes in the current period have a set of parent nodes;

[0031] The content verification unit is used to verify whether the node hash is consistent with the content and dual time tags of the original event;

[0032] The temporary root generation unit is used to generate a temporary periodic root when there is a node to be attached within the current period whose buffer time has not reached the allowable time window for backfilling;

[0033] The formal root upgrade unit is used to update the temporary periodic root to the formal periodic root when the buffer time of all nodes to be attached in the current period has reached the allowable time window for replenishment, and submits the periodic root to the on-chain anchoring and traceability verification module.

[0034] In one alternative: the on-chain anchoring and traceability verification module also includes a side branch evidence storage unit. The side branch evidence storage unit is used to receive nodes transferred into the side branch directed acyclic graph, generate a local closed subgraph in the side branch directed acyclic graph, generate a side branch anchoring record based on the local closed subgraph, and establish a hash pointer-based association with the main anchoring chain through the periodic root.

[0035] In one alternative approach: the on-chain anchoring and traceability verification module outputs the student's growth trajectory through the following steps: locating candidate event records based on the received student identifier and query interval through the event location unit;

[0036] The consistency verification unit verifies whether the original content hash, dual time tag, and source identifier of the candidate event record match the pre-stored hash value, pre-stored dual time tag, and pre-stored source identifier, respectively.

[0037] The attribution verification unit verifies whether the node corresponding to the candidate event record belongs to the current period's directed acyclic graph or a side branch directed acyclic graph.

[0038] Verify whether the cycle root of the node has been included in the main anchor chain by verifying the main chain inclusion verification unit.

[0039] The side branch association verification unit is used to verify that the hash pointer in the side branch anchor record is consistent with the hash proof certificate of the periodic root when the candidate event record corresponds to the node that is transferred into the side branch directed acyclic graph.

[0040] The trajectory reconstruction unit rearranges all verified candidate event records according to the event occurrence time, and outputs the student's growth trajectory.

[0041] In one alternative: the delay buffer module further includes a window adjustment unit, which is used to read the median and standard deviation of late arrival times within the sliding window, adjust the sliding window threshold for the next buffer cycle based on the median and standard deviation of late arrival times, obtain the number of reorderings triggered due to the inconsistency between the event reception order and the event occurrence order within a preset statistical period, and adjust the preset adjustment step size based on the number of reorderings.

[0042] In one alternative: the association determination module further includes a threshold adjustment unit, which is used to extract the determination accuracy, source credibility and delay penalty value from historical data and perform normalization processing;

[0043] The weighting coefficients for judgment accuracy, source credibility, and delay penalty value are determined based on historical data; the main chain batch threshold and side branch retention threshold are determined by weighted calculation based on the adjusted weighting coefficients.

[0044] In one alternative approach: the data structure of the original event includes at least the student identifier, event category identifier, event payload data, event occurrence time, system reception time, source credibility, and data source identifier;

[0045] The dual-timestamp standardization module writes the event occurrence time and the system reception time into the dual timestamps as the first timestamp and the second timestamp, respectively.

[0046] The beneficial effects of this invention are:

[0047] 1. The business data access module of this invention loads the original events sent by the teaching affairs system, student affairs system, teacher evaluation system, practical activity system, and IoT devices into standard data packets. The dual time tag standardization module simultaneously writes the event occurrence time and system reception time into the event record and calculates the late time. The association determination module then combines the student identifier, event category identifier, event payload data, source credibility, and data source identifier to determine the event association degree. The delay buffer module marks the event as a main batch candidate record, a supplementary candidate record, or a side branch candidate record based on the late time and event association degree and buffers it within a sliding window. The access phase-locked multi-source asynchronous data is placed in the time-series reconstructed buffer array, which reduces the logical causal misalignment rate caused by single-track reception and improves the topological convergence stability of the directed acyclic graph under real-time throughput conditions.

[0048] 2. The hash graph construction module of this invention prioritizes sorting the set of nodes to be solidified by the event occurrence time, and matches the event with a set of parent nodes based on the event category, student identifier, event occurrence time, and event processing decision. When the parent nodes are complete, a node hash containing the original event content, double time tags, and parent node set identifier is generated and written into the current periodic directed acyclic graph. For events that cannot be completed by the time window for backfilling, the parent nodes are transferred to the side branch of the directed acyclic graph. The Merkle tree evidence module constructs the periodic root for the closed node hash and generates a hash proof certificate. Then, the on-chain anchoring and traceability verification module appends the formal periodic root to the main anchor chain. This structure, while maintaining the non-writeability of the main anchor chain data, provides a topological path for event records that exceed the time window for backfilling to be attached to the side branch of the directed acyclic graph, ensuring the verification consistency and physical integrity of the historical evidence periodic roots.

[0049] 3. The parameter management process of this invention updates the sliding window threshold and adjustment step size for the next buffer cycle based on the median late time, standard deviation of late time, and number of rearrangements within the statistical period of all events within the sliding window. It also performs normalized weighted adjustment of the main chain batching threshold and side branch retention threshold based on the judgment accuracy, source credibility, and latency penalty value, while keeping the main chain batching threshold higher than the side branch retention threshold. By adaptively adjusting the sliding window threshold and event diversion boundary, the isolated node rate of the main chain and the search degradation rate of the side branches are reduced, maintaining the throughput balance of distributed data nodes under dynamic latency disturbances. Attached Figure Description

[0050] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0051] Figure 1 This is a structural block diagram of the student growth trajectory storage and traceability system based on hash chain anti-tampering in the embodiments of this application. Detailed Implementation

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

[0053] Please see Figure 1 A student growth trajectory storage and traceability system based on hash chain anti-tampering includes the following modules connected in sequence: business data access module, dual time tag standardization module, association judgment module, delay buffer module, hash graph construction module, Merkle tree storage module, and on-chain anchoring and traceability verification module.

[0054] The business data access module is used to collect raw events, which include the event occurrence time, system reception time, and event payload data.

[0055] The dual-time-tag standardization module is used to determine the late time based on the event occurrence time and system reception time of the original event, and generate event records with dual timestamps;

[0056] The correlation determination module is used to determine the correlation degree of events based on event payload data, and to generate event handling decisions based on the lateness time and event correlation degree.

[0057] The delay buffer module is used to buffer event records within a sliding window. When the window expires, the current batch meets the preset closing conditions, or the capacity limit is reached, the batch is sealed and the set of nodes to be solidified is output.

[0058] The hash graph construction module is used to sort the set of nodes to be solidified by event occurrence time and event processing decision, generate node hashes for events that match the parent node set according to preset dependencies, construct the directed acyclic graph of the current period, and output the hashes of closed nodes.

[0059] The Merkle tree notarization module is used to generate a periodic root and a hash proof certificate based on the hash of the closed node; the on-chain anchoring and traceability verification module is used to append the periodic root to the main anchoring chain record, and output the student's growth trajectory based on the main anchoring chain, the periodic root, the current periodic directed acyclic graph, and the hash proof certificate when a traceability request is received.

[0060] The correlation determination module includes a main batch decision unit, which generates a main batch entry decision when the event correlation is greater than or equal to the main batch entry threshold and the lateness time is less than or equal to the sliding window threshold.

[0061] The backfill determination unit is used to generate a backfill decision when the event correlation is greater than or equal to the main chain batch threshold and the sliding window threshold is less than the late time, and the late time is less than or equal to the product of the late amplification coefficient and the sliding window threshold; or when the side branch retention threshold is less than or equal to the event correlation and the event correlation is less than the main chain batch threshold, and the late time is less than or equal to the product of the late amplification coefficient and the sliding window threshold.

[0062] The side branch determination unit is used to generate a side branch diversion decision when the late arrival time is greater than the product of the late arrival amplification factor and the sliding window threshold, or when the event correlation is less than the side branch retention threshold.

[0063] The exception alarm unit is used to generate a discard alarm decision when the field integrity and format validation of the event record fails.

[0064] The correlation determination module outputs the main batch entry decision, the replenishment entry decision, the side branch diversion decision, or the discard alarm decision as event processing decisions. Among them, the main batch entry threshold, the side branch retention threshold, the late arrival amplification coefficient, and the sliding window threshold are all preset configuration parameters, and the main batch entry threshold is greater than the side branch retention threshold, and the late arrival amplification coefficient is greater than 1.

[0065] The delay buffer module includes: a candidate marking unit, which receives event records with dual time tags, marks the event records as main batch candidate records, supplementary candidate records or side branch candidate records based on event processing decisions, and stores them in the buffer pool;

[0066] The batch sealing trigger unit is used to trigger a batch sealing command when the waiting time in the buffer pool reaches the sliding window threshold, or the event records in the current batch meet the preset closing conditions, or the number of event records in the batch reaches the capacity limit.

[0067] The set output unit is used to output the event records that meet the closure conditions as a set of nodes to be solidified after receiving the sealing instruction, and to retain the event records that do not meet the closure conditions to the next round of buffer pool.

[0068] This implementation method is deployed in a university data middleware server cluster. The server cluster includes at least an access server, a processing server, an evidence storage server, and a query server. Event records are transmitted between the servers through a unified event bus.

[0069] The original event is configured as a data record sent by the academic affairs system, student affairs system, teacher evaluation system, practical activity system, IoT check-in device, dormitory access control device or psychological counseling business terminal. This data record shall at least contain student identifier, event category identifier, event payload data, event occurrence time, system reception time, source credibility and data source identifier;

[0070] Event payload data is the specific content of the corresponding business event, such as score, check-in result, activity participation information, reward and punishment explanation or evaluation text;

[0071] The event occurrence time indicates the time attribute of the business event, and the system reception time indicates the time attribute of the data platform receiving the business event. Together, they reflect the actual occurrence order and system entry order under the condition of multi-source asynchronous access.

[0072] The business data access module runs on the access server and is used to receive raw events from the interfaces of various business systems or device reporting interfaces.

[0073] During access, the fields are first loaded according to the preset data structure, the same event is organized into a data message in a unified format, and the data message is sent to the dual time tag standardization module.

[0074] In this step, the access server does not change the original event payload, but only completes the protocol conversion and field merging, thereby ensuring that the content corresponding to the subsequent hash processing is consistent with the original business record;

[0075] The dual-time-stamp standardization module runs on the processing server. After receiving the original event, it reads the event occurrence time and the system reception time from the event and calculates the delay time.

[0076] The delay time is calculated using the following formula to reflect the degree of lag in the event's entry into the system relative to its actual occurrence order:

[0077]

[0078] In the formula, Late arrival time represents the time difference due to system reception delay; This refers to the system receiving time. The time when the event occurred;

[0079] The standardization module writes the event occurrence time into the first timestamp, the system reception time into the second timestamp, and combines the late time with these two timestamps to form a dual timestamp.

[0080] The processing server concatenates the original event field with the double time tags to form an event record and then outputs it to the association determination module;

[0081] The dual time tagging mechanism configuration enables subsequent sorting to be performed based on the event occurrence time and reception time, preventing trajectory timing distortion caused by network delays or offline retransmission.

[0082] The association determination module also runs on the processing server; the event association degree refers to the strength of the association between a certain event and the existing event sequence in the student's growth process. It is used to reflect whether the event is suitable to enter the current main batch, whether it should be processed as a supplementary event, or whether it should be separated and retained as a side event.

[0083] During the association determination, the processing server extracts the student identifier, event category identifier, event payload data, source credibility, and data source identifier from the event record, and checks whether the event can form a reasonable sequential relationship with the existing events in the current buffer period according to the preset dependency relationship;

[0084] For example, there is usually a sequential dependency between the activity participation records and subsequent evaluation records of the same student, and there may also be a dependency within the same category based on the time of event occurrence;

[0085] The processing server stores a square matrix in its memory, representing the total number of event categories of a preset order, as an event dependency mapping matrix. The row and column indices of this matrix correspond to preset event category identifiers. The processing server retrieves the mapping values ​​from the matrix by extracting the current event category to be determined and comparing it with existing event categories.

[0086] when Furthermore, if the student identifiers match in a two-way comparison, it is determined to be a strong pre-dependency, and the base score is assigned to the first extreme value. ;

[0087] when Furthermore, if the student identifiers are consistent, it is determined to be a weak dependency, and the base score is assigned to the second extreme value. ;when If the student identifiers do not match, it is determined that there is no dependency relationship, and the base score is assigned. The server uses this information to establish an event correlation, which is then combined with the lateness time and fed into the decision-making logic.

[0088] To prevent type drift in the judgment of events whose relevance is greater than or equal to the main chain batch threshold and whose lateness time is within the preset conversion range, the system introduces relevance attenuation preprocessing for events in this specific range before performing the above judgment. That is, the original relevance is equivalently down-converted to a range greater than or equal to the side branch retention threshold and less than the main chain batch threshold according to the lateness exceedance, and the attenuated relevance is limited to a range when outputting.

[0089] If the correlation after decay is still greater than or equal to the main chain batch threshold, it is lowered to the specified boundary value obtained by subtracting the first fixed step size from the main chain batch threshold; if the conversion result is lower than the side branch retention threshold, it is raised to the side branch retention threshold, thereby ensuring that such events can be stably accepted by the backfill decision unit.

[0090] The specific conversion rule uses the following formula to perform correlation attenuation processing:

[0091] Attenuation ratio calculation formula:

[0092]

[0093] Formula for calculating the correlation after attenuation:

[0094]

[0095] In the formula, This refers to the attenuation ratio; The time of delay for the event; The threshold value for the current sliding window; Late amplification factor ( ), Indicates the maximum allowable replenishment range; The correlation after attenuation; The original correlation degree; The threshold for main chain batch input; Preserve a threshold for lateral branches;

[0096] Field integrity checks should at least include checking for missing student identifiers, missing event category identifiers, and the existence of fields required for dual time labels.

[0097] Format validation includes at least whether the time field is parsable and whether the payload field meets the preset encoding format; the correlation determination module outputs not only an abstract score, but also event processing decisions that can directly drive subsequent buffering and networking.

[0098] The delay buffer module runs in the buffer service process of the processing server; the sliding window is a buffer period parameter used to wait for late events to complete the correlation. It is not a fixed point in time, but moves continuously according to the system cycle.

[0099] After receiving the event handling decision, the candidate marking unit marks the event record as a main batch candidate record, a supplementary candidate record, or a side branch candidate record, and writes it into the buffer pool.

[0100] The buffer pool can be implemented by combining a memory queue and a persistent table, where the memory queue is used for fast processing in the current cycle, and the persistent table is used for record recovery after system restart;

[0101] The configuration structure of the buffer pool allows the system to delay the writing of the final evidence structure when it detects an IoT device or departmental interface retransmission event, and to reserve a topology waiting interval for the reconstruction of the time sequence relationship within the sliding window;

[0102] The batch sealing trigger unit continuously monitors the buffer pool status; the preset closure condition means that the preset dependencies of each event in the current batch have been met, and there are no missing parent node events in the current period; the batch sealing command is triggered when any of the following conditions are met:

[0103] First, the waiting time in the buffer pool reaches the sliding window threshold;

[0104] Secondly, the events within the current batch have met the preset closure conditions;

[0105] Third, the number of events in a batch reaches the capacity limit; the capacity limit is used to limit the processing scale of a single network setup and hash calculation, so as to avoid the excessive number of records in a single cycle from affecting server throughput;

[0106] After receiving the sealing instruction, the set output unit organizes the event records that have met the closure conditions into a set of nodes to be solidified and outputs it to the hash graph construction module; the event records that have not yet met the closure conditions are retained in the next round of buffer pool;

[0107] To prevent the batch calculation process from blocking the access of real-time asynchronous data, the buffer pool adopts a dual-cursor ring buffer structure for concurrent flow control;

[0108] When a batch is sealed, the collection output unit locks the read cursor boundary of the current batch and executes the solidified output by an asynchronous thread, while the candidate marker unit continues to advance the write cursor to accept new events without blocking.

[0109] For events that do not meet the closure condition and are retained to the next round, the buffer module adds a lifecycle count attribute to them. If the count flips with the cycle and exceeds the preset maximum lifecycle, it is transferred as an unclosed orphan data to an independent persistent exception data storage table to avoid continuous memory overflow.

[0110] The persistent abnormal data storage table is used to persistently store isolated events that cannot participate in the network due to the lack of a parent node for a long time. The data in this table will no longer participate in the construction of the hash map in the current and subsequent buffer cycles, and will only be used as a system abnormal record for back-end auditing and manual investigation.

[0111] When the business data access module prioritizes receiving the second event record, and the first event record that occurred before the record is not synchronized due to network delay, if the association determination module determines that the second event record has a prior dependency on the missing first event record and the late time meets the sliding window threshold condition, then it marks it as a candidate record for replenishment and stores it in the buffer pool.

[0112] Afterwards, the activity record was uploaded to the system. The standardization module calculated that although the record was received late, the event occurred earlier. The correlation determination module calculated that the correlation between the activity record and the evaluation record was greater than the preset correlation threshold.

[0113] After the waiting conditions in the buffer pool are met, the batching trigger unit issues a batching command, and the collection output unit sends the activity record and evaluation record to the next module together according to the reconstructible relationship; the trajectory reconstruction unit verifies and rearranges the events to be solidified based on the dual time tags, and reconstructs the growth trajectory based on the physical time series.

[0114] The system extracts the event occurrence time and the system reception time, and combines the lateness time with the event correlation to output a traffic distribution decision, thereby realizing the structured filtering and tamper-proof solidification of asynchronous access data.

[0115] This data processing mechanism is used to reduce the logical timing misalignment rate caused by single-order reception and outputs a standardized set of nodes to be solidified to the hash graph construction module.

[0116] In a preferred embodiment of the present invention, the hash graph construction module includes: a sorting unit, used to sort the event records in the set of nodes to be solidified according to the event occurrence time and event processing decision;

[0117] The parent node matching unit is used to filter the set of parent nodes for each sorted event record according to a preset dependency relationship;

[0118] The attachment storage unit is used to temporarily store the corresponding event record as a node to be attached when not all parent nodes are matched and the late time of the event record is less than or equal to a preset recovery allowable time window.

[0119] The node hash generation unit is used to generate node hashes when all parent nodes are available and write them into the current period's directed acyclic graph.

[0120] The side branch transfer unit is used to transfer the corresponding node into the side branch directed acyclic graph when the late time of the event record exceeds the allowable time window for backfilling and all parent nodes cannot be matched.

[0121] The Merkle tree evidence storage module includes: a dependency verification unit, used to verify whether all nodes in the current period have a set of parent nodes;

[0122] The content verification unit is used to verify whether the node hash is consistent with the content and dual time tags of the original event;

[0123] The temporary root generation unit is used to generate a temporary periodic root when there is a node to be attached within the current period whose buffer time has not reached the allowable time window for backfilling;

[0124] The formal root upgrade unit is used to update the temporary periodic root to the formal periodic root when the buffer time of all nodes to be attached in the current period has reached the allowable time window for replenishment, and submit the periodic root to the on-chain anchoring and traceability verification module.

[0125] The on-chain anchoring and traceability verification module also includes a side branch evidence storage unit, which is used to: receive nodes transferred into the side branch directed acyclic graph, generate a local closed subgraph in the side branch directed acyclic graph; generate side branch anchoring records based on the local closed subgraph; and establish a hash pointer-based association with the main anchoring chain through the periodic root.

[0126] The on-chain anchoring and traceability verification module outputs the student's growth trajectory through the following steps:

[0127] The event location unit is used to locate candidate event records based on the received student identifier and query range;

[0128] The consistency verification unit is used to verify whether the original content hash, dual time tags, and source identifier of the candidate event record match the pre-stored hash value, pre-stored dual time tags, and pre-stored source identifier, respectively.

[0129] The attribution verification unit is used to verify whether the node corresponding to the candidate event record belongs to the current period's directed acyclic graph or a side branch directed acyclic graph.

[0130] The main chain inclusion verification unit is used to verify whether the period root where the node is located has been included in the main anchor chain.

[0131] The side branch association verification unit is used to verify that the hash pointer in the side branch anchor record is consistent with the hash proof certificate of the periodic root when the node corresponding to the candidate event record is transferred into the side branch directed acyclic graph.

[0132] The trajectory reconstruction unit is used to rearrange all verified candidate event records according to the event occurrence time and output the student's growth trajectory.

[0133] This implementation method follows the output of the aforementioned set of nodes to be solidified, and focuses on explaining the formation of the current periodic directed acyclic graph, the generation of the periodic root, the evidence storage of side branches, and the source verification during querying.

[0134] The relevant functions are mainly deployed on the evidence storage server and the query server. The hash graph construction module and the Merkle tree evidence storage module run on the evidence storage server, while the on-chain anchoring and traceability verification module runs on the query server and the anchoring service process.

[0135] The current cycle directed acyclic graph refers to a directed acyclic graph structure with events as nodes and dependencies as edges within a buffer cycle, used to represent the sequential relationships between student growth events.

[0136] The parent node set is a group of preceding nodes that, in relation to a certain event, should logically precede that event and be referenced or depended upon by that event;

[0137] A lateral directed acyclic graph is a graph structure created separately for events that are too late or cannot complete the parent node within the allowed time window for completion.

[0138] The hash graph construction module sorts the set of nodes to be solidified by the sorting unit; the sorting criteria include at least the event occurrence time and the event handling decision; if there is a difference in the occurrence time of two events, they are sorted first by the event occurrence time.

[0139] If the events occur at the same time under the precision used by the system, they are then sorted according to the event handling decision order to ensure that the main batch candidates, backup candidates, and side branch candidates have a stable processing order when the network is formed.

[0140] To reconstruct the real business time sequence logic, the sorting unit uses the event occurrence time as the primary sorting criterion to sort the set of nodes to be solidified;

[0141] After sorting, the parent node matching unit filters the set of parent nodes for each event in turn; the preset dependency relationship is a matching rule determined based on the event category, student identifier, event occurrence time and event handling decision;

[0142] For example, for the same student, the activity record can serve as the parent node of the activity evaluation record, and the event in the previous stage can serve as the parent node of the same type of event in the next stage.

[0143] Parent-child relationships are not established between different students, but events from different data sources that belong to the same student and the same growth process can be associated when the rules are met.

[0144] The evidence storage server checks each sorted event for its referential preceding events and writes the preceding nodes that satisfy the dependency relationship into the set of parent nodes of the event.

[0145] If the parent node matching unit finds that an event has not yet been matched with all parent nodes, the temporary storage unit continues to determine whether the late time of the event is less than or equal to the allowable time window for replenishment.

[0146] The backfill allowable time window is another periodic parameter used to control the waiting time for the parent node to complete the backfill. It works in conjunction with the aforementioned sliding window: the sliding window mainly determines the timing of buffering and sealing, while the backfill allowable time window mainly determines how long the waiting time is allowed during the networking phase.

[0147] If the lateness is still within the allowed time window for remediation, the event is temporarily stored as a node to be attached, and the final node hash is not generated immediately. The attachment temporary storage unit avoids triggering write-back conflicts in the historical graph structure by temporarily storing the node to be attached and waiting for subsequent parent node matching.

[0148] When all parent nodes are available, the node hash generation unit generates a node hash for the event. The input to the node hash includes at least the original event content, two timestamps, and parent node set identifiers. The hash is concatenated in a fixed byte order: first timestamp + second timestamp + original event content bitstream + parent node set identifiers arranged in ascending order of node generation time. Then, a 256-bit secure hash algorithm is performed on the concatenated byte array to obtain the node hash value of the event.

[0149] To provide a comparison benchmark for consistency verification in the subsequent traceability stage, the evidence storage server extracts the original event content of the event simultaneously when generating the node hash, calculates the original content hash independently, and solidifies it together with the dual time tags and source identifier into the index library, which serve as the pre-stored hash value, pre-stored dual time tags, and pre-stored source identifier, respectively.

[0150] Since the hash input already contains the event occurrence time, system reception time, and parent node relationships, any changes to the original content, timestamps, or dependencies will cause changes to the node hash.

[0151] After the node hash is generated, the node is written into the current period's directed acyclic graph. If the event's lateness time has exceeded the allowable time window for backfilling and it still cannot match all parent nodes, the side branch transfer unit will transfer the event into the side branch directed acyclic graph.

[0152] While maintaining the logical timing of the main graph, the side branch transfer unit writes the overdue late event into the side branch path to complete the topology connection.

[0153] After receiving the closed node hashes in the directed acyclic graph of the current period, the Merkle tree evidence storage module begins to construct the period root. The period root is the root hash value obtained by performing Merkle tree calculations on a set of node hashes within the period, and is used to represent the evidence storage result for the entire period.

[0154] The hash proof certificate is the proof path data used to backtrack from a leaf node to the root of the cycle; the dependency verification unit in the Merkle tree proof module first verifies whether all nodes in the current cycle have a set of parent nodes;

[0155] If a node is in a pending connection state but its parent node has not yet been determined, then the node will not be included in the calculation of the formal root.

[0156] The content verification unit verifies whether the node hash is consistent with the original event content and the dual time tags. The verification method is to re-extract the original event and dual time tags corresponding to the node, recalculate the hash in the same order when the node hash was generated, and compare it with the pre-stored node hash. If the two are consistent, it means that no content change has occurred during the intermediate processing.

[0157] If there are still nodes to be attached within the current cycle whose buffer time has not reached the allowable time window for replenishment, the temporary root generation unit generates a temporary cycle root for the currently closed node.

[0158] Temporary roots are only used for saving intermediate states within a cycle and are not used as formal additions to the main anchor chain. When the buffer time of all nodes to be attached reaches the allowable time window for replenishment, the formal root upgrade unit re-determines the final set of nodes to be included in this cycle, updates the temporary cycle root to the formal cycle root, and generates the corresponding hash proof certificate and submits it to the on-chain anchoring and traceability verification module.

[0159] The temporary root generation unit and the formal root upgrade unit work together in a preset time sequence to complete node dependency completion and main chain anchoring.

[0160] After receiving the official periodic root, the on-chain anchoring and traceability verification module appends the periodic root to the main anchoring chain. The main anchoring chain saves each periodic root in an append-only manner, so the new periodic root is only appended to the existing record and will not overwrite or modify the existing periodic root.

[0161] For nodes that are transferred into a lateral directed acyclic graph, the lateral evidence storage unit first searches for a locally closed subgraph in the lateral graph; a locally closed subgraph refers to a subset of nodes that can self-consistently represent a supplementary event relationship within the lateral graph.

[0162] The specific logic of the side branch evidence unit to find local closed subgraphs is as follows: Starting from the node in the current side branch graph, perform a depth-first reverse traversal along the preset dependency relationship, and check whether all the parent nodes of each node on the traversal path already exist in the existing graph structure corresponding to the historical period root of the current side branch graph or the main anchor chain.

[0163] If the set of connected nodes obtained by traversal no longer contains events to be attached that are missing parent nodes, then the set is determined to be a locally closed subgraph.

[0164] To avoid computational overload during graph traversal caused by a large number of incomplete records in extreme packet loss scenarios, the side branch evidence storage unit is configured with a dynamic search depth truncation threshold during reverse traversal. ,in Defined as the maximum event capacity limit constrained by the current sliding window threshold. This is multiplied to cover extreme concurrency scenarios;

[0165] If the recursive search depth of the current graph structure If the complete parent node chain is still not found, the call stack is immediately interrupted and the search degradation strategy is triggered: the current traversal process is interrupted, and the unclosed subset of nodes is marked as dangling and stored in the suspend queue.

[0166] The system performs periodic secondary matching by polling tasks at regular intervals. If the suspension time exceeds the maximum retention window, a virtual root node containing the source tracing warning flag is generated and attached to the head of the current subset to complete the physical closure.

[0167] The side branch evidence storage unit generates side branch anchoring records based on the locally closed subgraphs that have been successfully closed or forcibly closed, and then establishes a hash pointer-based association with the main anchoring chain through the periodic root;

[0168] Hash pointer-based association refers to the side branch anchoring record being configured with hash reference data pointing to the target periodic root or the corresponding proof path;

[0169] Side branch anchoring records establish historical topological relationships between side branch events and the existing main graph by referencing data through hashing.

[0170] The query phase is executed by the query server; after receiving the student identifier and query range, the event location unit first finds the candidate event record corresponding to the student and query range in the index; the query range can be a preset business range or the time range submitted by the query client.

[0171] The consistency verification unit verifies whether the original content hash, dual time tags, and source identifier of each candidate event are consistent with the pre-stored values. During verification, the query server rereads the original content of the candidate event and calculates the content hash, and then compares it with the pre-stored hash value.

[0172] At the same time, the double timestamps and source identifiers in the event record are compared to the values ​​retained during evidence preservation; only if all three match will the event proceed to the next step.

[0173] The attribution verification unit determines whether the node corresponding to the event belongs to the current period's directed acyclic graph or a lateral directed acyclic graph; if the node is neither in the main graph nor in the lateral graph, then the event does not participate in trajectory reconstruction.

[0174] The main chain inclusion verification unit further verifies whether the period root of the period in which the node is located has been included in the main anchor chain, that is, checks whether the period root exists in the appended main chain record;

[0175] If the candidate event is a side event, the side association verification unit continues to compare whether the hash pointer in the side anchor record is consistent with the hash proof certificate of the period root to confirm that the correspondence between the side event and the target period root has not been destroyed.

[0176] After completing the above verification, the trajectory reconstruction unit rearranges all verified candidate event records according to the event occurrence time and outputs the student's growth trajectory; if the event occurrence time is the same, it can be arranged according to the order of their dependencies in the graph structure to ensure that the reconstruction result is consistent with the business logic.

[0177] Taking the multi-source asynchronous access scenario as an example: A student's activity record is uploaded outside the period to which the main graph belongs, and the upload time has exceeded the allowed time window for re-upload. The system can no longer use it as a parent node to write back to the existing evaluation node.

[0178] At this point, the side branch transfer unit puts the activity record into the side branch graph, the side branch evidence storage unit generates a local closed subgraph and a side branch anchoring record, and associates the record with the target periodic root through a hash pointer;

[0179] When querying the student's growth trajectory later, the event localization unit simultaneously retrieves the evaluation records in the main graph and the activity records in the side graph;

[0180] After all consistency verification, attribution verification, main chain inclusion verification, and side branch association verification are passed, the trajectory reconstruction unit outputs the growth trajectory of the activity first and the evaluation later according to the event occurrence time; while maintaining the integrity of the main anchor chain data, the trajectory reconstruction unit outputs the source tracing results according to the event occurrence time.

[0181] The technical function of this implementation is to entrust the actual time sequence and the non-rewritable evidence storage to the directed acyclic graph and the main anchor chain respectively. The former is responsible for expressing the dependency relationship, and the latter is responsible for storing the immutable periodic root.

[0182] The on-chain anchoring and traceability verification module uses side graphs and hash pointers to build associated paths for late events that exceed a preset threshold, in order to complete the topology connection of asynchronous data retransmission.

[0183] In a preferred embodiment of the present invention, the delay buffer module further includes a window adjustment unit, which is used to: read the median of late arrival time and the standard deviation of late arrival time within the sliding window;

[0184] Adjust the sliding window threshold for the next buffer cycle based on the median and standard deviation of late arrival times; obtain the number of rearrangements triggered by the inconsistency between the event reception order and the event occurrence order within the preset statistical period, and adjust the preset adjustment step size based on the number of rearrangements;

[0185] The association determination module also includes a threshold adjustment unit, which is used to: extract the determination accuracy, source credibility and delay penalty value from historical data and perform normalization processing;

[0186] The weighting coefficients for judgment accuracy, source credibility, and delay penalty value are determined based on historical data; the main chain batch entry threshold and side branch retention threshold are determined by weighted calculation based on the adjusted weighting coefficients.

[0187] The data structure of the original event should include at least the student identifier, event category identifier, event payload data, event occurrence time, system reception time, source credibility, and data source identifier;

[0188] The dual-timestamp standardization module writes the event occurrence time and the system reception time into the dual timestamps as the first timestamp and the second timestamp, respectively.

[0189] This implementation mainly explains the adjustment methods of the sliding window threshold and the association determination threshold, and further explains the writing method of the original event data structure and the dual time tag.

[0190] The relevant functions are executed by the parameter management process on the processing server. The parameter management process shares the configuration table with the aforementioned standardization module, association judgment module and delay buffer module. The median late time is the median statistic after arranging the late times of all events in the current sliding window according to their magnitude, and is used to reflect the delay level of most events.

[0191] The standard deviation of late arrival time is used to reflect the dispersion of late arrival time distribution within the same window; the number of reorderings refers to the number of times the sorting adjustment is triggered within the statistical period because the order of event reception is inconsistent with the order of event occurrence; the delay penalty value is used to reflect the negative impact of late arrival time on the association determination result;

[0192] The data structure of the original event should include at least the student identifier, event category identifier, event payload data, event occurrence time, system reception time, source credibility, and data source identifier;

[0193] After the server receives the raw events, it first writes them into the event cache table in a unified field order; the student identifier is used to classify the event into the corresponding student's growth trajectory; the event category identifier is used to distinguish business types such as grades, practice, evaluation, activities, and rewards and punishments; the event payload data is used to store specific business content.

[0194] The event occurrence time and system reception time are used together to calculate the delay time; source credibility is used to indicate the reliability of the data source; data source identifier is used to record which business system or device interface the event originated from;

[0195] The dual-time tag standardization module writes the event occurrence time into the first timestamp, the system reception time into the second timestamp, and attaches the late time to the same event record, so as to preserve both the actual occurrence order and the system reception order in the subsequent sorting, judgment, buffering and tracing stages.

[0196] After each buffer cycle, the window adjustment unit reads all late time samples in the sliding window and calculates the median and standard deviation of late time. If the median late time increases, it indicates that the entry delay of most events into the system has increased, and the sliding window threshold of the next buffer cycle needs to be appropriately widened.

[0197] If the median delay time decreases, it indicates that most events in the system can arrive relatively quickly, and the sliding window threshold can be appropriately tightened to reduce waiting time;

[0198] If the standard deviation of the late arrival time is greater than the first preset dispersion, the window adjustment step size is limited to the first set range; if the standard deviation of the late arrival time is less than or equal to the first preset dispersion, the sliding window threshold adjustment is performed using the second set range.

[0199] The specific update logic is as follows: when the median late time is greater than the current sliding window threshold, the next sliding window threshold is equal to the current threshold plus the adjustment step size;

[0200] When the median lateness time is less than half of the current sliding window threshold and the standard deviation of lateness time is less than the preset lower limit, the threshold of the next sliding window is equal to the current threshold minus the adjustment step size; otherwise, the threshold of the next sliding window remains unchanged.

[0201] The parameter management process updates the sliding window threshold for the next buffer cycle accordingly; this update only applies to the next buffer cycle and does not apply to already sealed event records, so it will not change the data that has already entered the subsequent evidence storage process.

[0202] The normalization process specifically employs range standardization, which involves linearly mapping each indicator to its minimum and maximum observed values ​​within the historical statistical period. The mapping result is then limited to a range of values. Its normalization result can be expressed as:

[0203]

[0204] in, For the normalized result, For each indicator, the original observed value, This represents the minimum observed value of this indicator within the historical statistical period. This represents the maximum observed value of this indicator within the historical statistical period; when the maximum value is equal to the minimum value, the normalization result of this indicator is recorded as 0.

[0205] Range standardization is used to unify the data units for determining accuracy, source reliability, and delay penalty values; if the adjusted sliding window threshold is less than the preset minimum threshold, the preset minimum threshold is used to avoid non-positive waiting windows after the threshold is reduced.

[0206] To prevent high-frequency oscillations and overflows in the sliding window threshold, the window adjustment unit dynamically constrains and adjusts the step size based on the number of rearrangements within the statistical period. When the number of rearrangements is greater than the first preset frequency, the window threshold is updated using the preset maximum step size. When the number of rearrangements is less than or equal to the first preset frequency, the threshold is updated using the preset baseline step size. The adjustment step size refers to the allowable change in each round of window threshold adjustment. The system presets a baseline step size and a maximum step size for this purpose.

[0207] The specific step size adjustment rules are as follows: when the number of rearrangements within the statistical period is greater than or equal to the preset number of alarms, the adjustment step size is the maximum step size; when the number of rearrangements is greater than 0 and less than the preset number of alarms, the adjustment step size is the baseline step size.

[0208] When the number of rearrangements is 0, the adjustment step size is half of the baseline step size; the window adjustment unit combines the median of late arrival time, standard deviation and number of rearrangements to constrain the sliding window parameters, and controls the batch triggering delay and event rearrangement rate within the preset performance threshold range;

[0209] The threshold adjustment unit is used to update the main chain batch threshold and the side branch retention threshold; the parameter management process extracts the judgment accuracy, source credibility and delay penalty value from historical data, and performs normalization processing first;

[0210] The accuracy rate is calculated from historical samples using a weighted summation method. The specific mapping rule is as follows: the proportion of isolated nodes within the statistical period is extracted as the main judgment validity, the proportion of forced virtual root attachment of side branches is extracted as the supplementary judgment validity, and the parent node reset rate caused by rearrangement is extracted as the side branch judgment validity. These are multiplied by the first preset weight, the second preset weight, and the third preset weight, and then summed. The sum of the first preset weight, the second preset weight, and the third preset weight is limited to 1. The source credibility is directly derived from the source credibility field in the original event.

[0211] The delay penalty value is obtained by mapping the position of the late time relative to the sliding window threshold and the allowable time window for make-up through a linear piecewise function: when the late time is less than or equal to the sliding window threshold, the delay penalty value is recorded as 0;

[0212] When the late time is between the sliding window threshold and the allowable recovery time window, the delay penalty value is equal to the difference between the current late time and the sliding window threshold, divided by the difference between the allowable recovery time window and the sliding window threshold; when the late time is greater than the allowable recovery time window, the delay penalty value is recorded as 1.0.

[0213] The normalization process specifically employs range standardization, mapping the three indicators mentioned above to... Within a unified dimension range, subsequent weighted calculations will not become unbalanced due to different original value ranges;

[0214] Based on this, the threshold adjustment unit determines the weight coefficients of judgment accuracy, source credibility and delay penalty value according to historical data; the specific weight adjustment logic is as follows: at the end of each statistical period, calculate the rate of isolated nodes in the main batch that are transferred to the persistent anomaly queue due to missing dependencies, and the rate of side branch hanging due to search degradation caused by not finding a local closed subgraph.

[0215] When the false entry rate of the main batch exceeds the preset tolerance limit, the current weight of the delay penalty value is increased by the preset weight adjustment step size, and the current weight of the judgment accuracy is decreased by the same step size to strengthen the effect of the delay penalty.

[0216] When the false entry rate of side branches exceeds the preset tolerance limit, the current weight of the source credibility is added to the preset weight adjustment step size, and the current weight of the judgment accuracy is subtracted by the same step size, thereby increasing the chance of high credibility events being accepted by the main batch.

[0217] After performing dynamic weight adjustment, in order to ensure parameter convergence and prevent divergence, the parameter management process performs forced normalization and lower limit constraint on the three weight coefficients: the three adjusted weights are treated as intermediate values ​​and summed. If the sum deviates from 1.0, it is scaled proportionally according to the relative proportion of each intermediate value so that the sum is strictly equal to 1.0.

[0218] Monitor whether any weight after scaling is lower than the preset safety lower limit constraint value, such as 0.1. If it is, force it to be locked as the constraint lower limit and compensate the difference to the other weights that have not exceeded the limit by means of reverse proportional distribution.

[0219] The parameter management process substitutes the final weight coefficients after normalization and boundary constraint processing into the weighted calculation to obtain the new main chain batch threshold and side branch retention threshold, and writes them back to the configuration table.

[0220] Before writing back to the configuration table, the system will also impose upper and lower bound constraints on both to ensure that the main chain batch threshold is always greater than the side branch retention threshold.

[0221] If the weighted calculation results show overlapping intervals, the main chain threshold is set to the upper limit of the interval extreme value, and the side branch threshold is set to the lower limit of the interval extreme value to perform boundary separation.

[0222] The specific quantification of the main chain batch entry threshold is calculated using the following weighted formula:

[0223]

[0224] In the formula, This is the dynamically adjusted main chain batch threshold; The preset main chain baseline value; The normalized accuracy rate; The weighting coefficients for determining accuracy; The source credibility after normalization; This is a weighting coefficient for the credibility of the source; This is the normalized delay penalty value; The weighting coefficient for the delay penalty value;

[0225] The calculation structure for the side branch retention threshold is similar, but its baseline value is smaller and it places more emphasis on source credibility weight; the specific quantification of the side branch retention threshold is calculated using the following weighted formula:

[0226]

[0227] In the formula, To retain a threshold for dynamically adjusted lateral branches, The preset lateral branch baseline value;

[0228] For example, when the accuracy rate is 0.8, the source credibility is 0.9, the delay penalty is 0.2, and the corresponding weights are 0.5, 0.3, and 0.2 respectively, and the main chain baseline value is 0.2, the dynamically adjusted main chain batch threshold is 0.2 + (0.8 × 0.5 + 0.9 × 0.3 - 0.2 × 0.2) = 0.83;

[0229] Through this structured decomposition and example, the calculation rules from historical feedback to dynamic thresholds are clarified; since the main chain batch threshold requirement is higher than the side branch retention threshold, the size relationship between the two is maintained after the update, thereby ensuring that the three processing paths of main batch, backfill, and side branch have stable boundaries.

[0230] The parameter management process reads the median, standard deviation and number of reordering times in real time, and dynamically updates the main chain batch threshold and side branch retention threshold based on source credibility and latency penalty value, so that the diversion decision can adaptively adapt to the latency characteristics of the current data source.

[0231] The original events are encapsulated with double time tags to determine the order of occurrence and receipt. The system dynamically optimizes the buffer and judgment parameters based on the historical late arrival distribution and source credibility to control the data distribution ratio of the main batch inbound chain, the supplementary inbound chain, and the side branch diversion.

Claims

1. A student growth trajectory storage and traceability system based on hash chain for tamper-proof protection, characterized in that, It includes the following modules that are connected in sequence: business data access module, dual time tag standardization module, association determination module, delay buffer module, hash graph construction module, Merkle tree evidence storage module, and on-chain anchoring and traceability verification module; The business data access module is used to collect raw events. The data structure of the raw events includes at least student identifier, event category identifier, event payload data, event occurrence time, system reception time, source credibility, and data source identifier. The event payload data includes at least one of the following: grade value, attendance result, activity participation information, reward and punishment description, or evaluation text. The dual-time-tag standardization module is used to determine the late time based on the event occurrence time and system reception time of the original event, and generate event records with dual timestamps; The association determination module is used to determine the association degree of an event based on the student identifier, event category identifier, event payload data, source credibility, and data source identifier, and to generate an event handling decision based on the lateness time and event association degree. The delay buffer module is used to buffer event records within a sliding window. When the window expires, the current batch meets the preset closing conditions, or the capacity limit is reached, the batch is sealed and the set of nodes to be solidified is output. The hash graph construction module is used to sort the set of nodes to be solidified by event occurrence time and event processing decision, generate node hashes for events that match the parent node set according to preset dependencies, construct the directed acyclic graph of the current period, and output the hashes of closed nodes. The Merkle tree notarization module is used to generate a formal periodic root and hash proof certificate based on the hash of the closed node; the on-chain anchoring and traceability verification module is used to append the periodic root to the main anchoring chain record, and output the student's growth trajectory based on the main anchoring chain, periodic root, current period directed acyclic graph and hash proof certificate when a traceability request is received. The hash graph construction module includes: a sorting unit, used to sort the event records in the set of nodes to be solidified according to the event occurrence time and event processing decision; The parent node matching unit is used to filter the set of parent nodes for each sorted event record according to a preset dependency relationship; The attachment storage unit is used to temporarily store the corresponding event record as a node to be attached when not all parent nodes are matched and the late time of the event record is less than or equal to a preset recovery allowable time window. The node hash generation unit is used to generate node hashes when all parent nodes are available and write them into the current period's directed acyclic graph. The side branch transfer unit is used to transfer the corresponding node into the side branch directed acyclic graph when the late time of the event record exceeds the allowable time window for backfilling and still cannot match all parent node closures.

2. The student growth trajectory storage and traceability system based on hash chain anti-tampering as described in claim 1, characterized in that, The correlation determination module includes a main batch decision unit, which generates a main batch entry decision when the event correlation is greater than or equal to the main batch entry threshold and the lateness time is less than or equal to the sliding window threshold. The backfill determination unit is used to generate a backfill decision when the event correlation is greater than or equal to the main chain batch threshold and the sliding window threshold is less than the late time, and the late time is less than or equal to the product of the late amplification coefficient and the sliding window threshold; or when the side branch retention threshold is less than or equal to the event correlation and the event correlation is less than the main chain batch threshold, and the late time is less than or equal to the product of the late amplification coefficient and the sliding window threshold. The side branch determination unit is used to generate a side branch diversion decision when the late arrival time is greater than the product of the late arrival amplification factor and the sliding window threshold, or when the event correlation is less than the side branch retention threshold. An anomaly alarm unit is used to generate a discard alarm decision when the field integrity and format verification of the event record fails. The association determination module outputs the main batch entry decision, the replenishment entry decision, the side branch diversion decision, or the discard alarm decision as event processing decisions. Among them, the main batch entry threshold, the side branch retention threshold, the late arrival amplification coefficient, and the sliding window threshold are all preset configuration parameters, and the main batch entry threshold is greater than the side branch retention threshold, and the late arrival amplification coefficient is greater than 1.

3. The student growth trajectory storage and traceability system based on hash chain anti-tampering as described in claim 2, characterized in that, The delay buffer module includes: a candidate marking unit, which receives event records with dual time tags, marks the event records as main batch candidate records, supplementary candidate records or side branch candidate records based on event processing decisions, and stores them in the buffer pool; The batch sealing trigger unit is used to trigger a batch sealing command when the waiting time in the buffer pool reaches the sliding window threshold, or the event records in the current batch meet the preset closing conditions, or the number of event records in the batch reaches the capacity limit. The set output unit is used to output the event records that meet the closure conditions as a set of nodes to be solidified after receiving the sealing instruction, and to retain the event records that do not meet the closure conditions to the next round of buffer pool.

4. The student growth trajectory storage and traceability system based on hash chain anti-tampering as described in claim 1, characterized in that, The Merkle tree evidence storage module includes: a dependency verification unit, used to verify whether all nodes in the current period have a set of parent nodes; The content verification unit is used to verify whether the node hash is consistent with the content and dual time tags of the original event; The temporary root generation unit is used to generate a temporary periodic root when there is a node to be attached within the current period whose buffer time has not reached the allowable time window for backfilling; The formal root upgrade unit is used to update the temporary periodic root to the formal periodic root when the buffer time of all nodes to be attached in the current period has reached the allowable time window for replenishment, and submit the formal periodic root to the on-chain anchoring and traceability verification module.

5. The student growth trajectory storage and traceability system based on hash chain anti-tampering as described in claim 1, characterized in that, The on-chain anchoring and traceability verification module also includes a side branch evidence storage unit, which is used to: receive nodes transferred into the side branch directed acyclic graph, generate a local closed subgraph in the side branch directed acyclic graph; generate side branch anchoring records based on the local closed subgraph; and establish a hash pointer-based association with the main anchoring chain through the formal periodic root.

6. The student growth trajectory storage and traceability system based on hash chain anti-tampering as described in claim 5, characterized in that, The on-chain anchoring and traceability verification module outputs the student's growth trajectory through the following steps: The event location unit is used to locate candidate event records based on the received student identifier and query range; The consistency verification unit is used to verify whether the original content hash, dual time tags, and source identifier of the candidate event record match the pre-stored hash value, pre-stored dual time tags, and pre-stored source identifier, respectively. The attribution verification unit is used to verify whether the node corresponding to the candidate event record belongs to the current period's directed acyclic graph or a side branch directed acyclic graph. The main chain inclusion verification unit is used to verify whether the formal cycle root where the node is located has been included in the main anchor chain. The side branch association verification unit is used to verify that the hash pointer in the side branch anchor record is consistent with the hash proof certificate of the formal periodic root when the candidate event record corresponds to the node that is transferred to the side branch directed acyclic graph. The trajectory reconstruction unit is used to rearrange all verified candidate event records according to the event occurrence time and output the student's growth trajectory.

7. The student growth trajectory storage and traceability system based on hash chain anti-tampering as described in claim 2, characterized in that, The delay buffer module also includes a window adjustment unit, which is used to: read the median of late times and the standard deviation of late times within the sliding window; Adjust the sliding window threshold of the next buffer cycle based on the median and standard deviation of the late arrival time; obtain the number of rearrangements triggered by the inconsistency between the event reception order and the event occurrence order within a preset statistical period, and adjust the preset adjustment step size based on the number of rearrangements.

8. The student growth trajectory storage and traceability system based on hash chain anti-tampering as described in claim 2, characterized in that, The association determination module also includes a threshold adjustment unit, which is used to: extract the determination accuracy, source credibility and delay penalty value from historical data and perform normalization processing; The weighting coefficients for judgment accuracy, source credibility, and delay penalty value are determined based on historical data; the main chain batch threshold and side branch retention threshold are determined by weighted calculation based on the adjusted weighting coefficients.

9. The student growth trajectory storage and traceability system based on hash chain anti-tampering as described in claim 1, characterized in that, The dual-timestamp standardization module writes the event occurrence time and the system reception time into the dual timestamps as the first timestamp and the second timestamp, respectively.

Images