Method and system for electronic file organization and single-set file transfer under four-dimensional environment

Abstract

The present invention relates to the technical field of electronic archive management, and discloses a method and system for electronic archive organization and single-set archive transfer under a four-quality environment. The method collects electronic files during the construction period; adds the electronic files to the approval process and performs electronic signatures to generate a file stream containing signature stream data; pre-archives and classifies the approved electronic files; uses a detection strategy based on an adversarial detection model to perform four-quality detection on the pre-archived files, the four qualities including authenticity, integrity, availability, and security; and registers the electronic files from the construction period that pass the detection into a database for archiving and transfer. This method improves the approval efficiency of engineering electronic files and enhances the integration of the electronic signature process with the project management system. The coverage of the four-quality detection is expanded to include defect types that are difficult to identify with traditional methods, such as time series anomalies and metadata logic errors, and can effectively reduce the risk of missed detection and misjudgment.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic archive management, and in particular to a method and system for electronic archive organization and single-set archive transfer under a four-dimensional environment. Background Art

[0002] In today's digital age, engineering and construction projects generate a vast amount of electronic documents, and their importance is growing. Traditional methods for engineering document management and file transfer often rely on manual operations, which are inefficient and prone to errors. These methods fail to meet the efficiency, accuracy, and security requirements of modern engineering projects. With the development of electronic signature technology, while some engineering projects have begun adopting electronic signatures for document signing, integration with project management systems remains limited. Interface compatibility issues between different platforms, inconsistent signature rules, and incomplete permission management have limited the application of electronic signatures in engineering document management, preventing them from fully realizing their advantages. When verifying electronic engineering documents, verifying their authenticity, integrity, availability, and security (collectively, the "four properties") is crucial. However, existing verification methods often suffer from incompleteness and low accuracy. Some methods only test a single aspect of electronic documents, failing to comprehensively assess their overall quality. Some verification technologies are inefficient when dealing with complex engineering document formats and large amounts of file data, making them inadequate for real-world project requirements. Furthermore, existing file transfer methods often rely on a dual-copy system, where both paper and electronic files coexist. This not only increases management costs but also easily leads to inconsistent file versions. Therefore, a solution that effectively integrates electronic signature technology, comprehensively verifies the four properties of electronic files, and implements a single-copy file transfer system is urgently needed to improve the overall efficiency and security of engineering electronic file management. To address these issues, existing technologies urgently need improvement. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a method and system for electronic file organization and single-set file transfer under a four-quality environment, so as to improve the degree of automation of electronic file management and ensure the comprehensiveness and accuracy of the four-quality detection.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] The method for electronic archive organization and single-set archive transfer under a four-quality environment includes: collecting electronic files during the construction period; adding electronic files to the approval process and electronically signing and sealing them to generate a file flow containing signature flow data; pre-archiving and classifying the electronic files that have completed approval; using a detection strategy based on an adversarial detection model to perform four-quality detection on the pre-archived files, and the four qualities include authenticity, integrity, availability and security; registering the electronic files during the construction period that pass the detection into the warehouse, and archiving and transferring them.

[0006] In the present invention, preferably, the adversarial detection model adopts an adversarial training strategy: creating a virtual signature node, generating adversarial flow data containing a forged certificate chain and an abnormal timestamp; constructing metadata conflict samples, including abnormal sample data with enumeration value out of bounds, format misplacement and repeated fields; mixing the adversarial flow data, abnormal sample data and real data in a preset proportion to generate a training set, and using the training set to perform adversarial training on the adversarial detection model multiple times until the adversarial detection model's misjudgment rate of the input data is lower than a threshold.

[0007] In the present invention, preferably, the detection strategy for the authenticity detection specifically includes: extracting the timestamp sequence in the file stream, constructing a time series feature vector through an LSTM network; comparing the actual signature node with the time window matching of the approval process; calculating the cryptographic strength index of the certificate chain, including key length, hash collision probability and signature timeliness; inputting the time series feature vector, time window matching and cryptographic strength index into the adversarial detection model after adversarial training, and outputting a credibility score.

[0008] In the present invention, preferably, the detection strategy for the integrity detection specifically includes: constructing a metadata topology graph and analyzing the logical dependencies between fields; detecting abnormal connection edges and missing nodes through a graph neural network; calculating the semantic similarity between the file entity and the metadata description; dynamically generating adversarial text containing randomly missing fields to verify the error rate of the adversarial detection model.

[0009] In the present invention, preferably, the detection strategy for the availability detection specifically includes: determining whether the format of the pre-archived electronic file meets the requirements of long-term storage; determining whether the metadata of the pre-archived electronic file is accessible and whether the metadata in the information package is readable.

[0010] In the present invention, preferably, the detection strategy for the security detection specifically includes: generating attack simulation data of data tampering attacks, signature forgery attacks and structure destruction attacks, mixing the attack simulation data with real data in a preset proportion and putting them into the training set of the adversarial detection model, training the adversarial detection model until the misjudgment rate of the adversarial detection model for the input data is lower than the threshold; using the adversarial detection model generated by gradient penalty optimization, and using the optimized adversarial detection model to perform security detection on electronic files.

[0011] In the present invention, preferably, it also includes dynamic threshold adjustment of the adversarial detection model: establishing a Bayesian probability distribution model of the four indicators based on historical detection data; calculating the KL divergence of the detection results of the current batch and the probability distribution in real time; when the divergence value exceeds the warning threshold, triggering the adversarial data generation module to update the training set; adjusting the weight coefficient of each detection dimension to form a closed-loop feedback and dynamic optimization of the adversarial detection model.

[0012] In the present invention, preferably, the electronic file package to be archived and transferred includes the encrypted original electronic file, the complete file stream with a time stamp, and the verification matrix.

[0013] In the present invention, preferably, a system for electronic file organization and single-set file transfer under a four-property environment is also proposed, including a receiving module, an approval module, an electronic signature module, an adversarial detection model, an electronic file storage module, and an archiving and handover module. The receiving module is used to receive electronic files during the construction period; the approval module initiates an approval process for the electronic files and performs step-by-step approval; the electronic signature module communicates with the contract lock platform to electronically sign the electronic files; extracts keywords from the electronic files that need to be signed, and determines the signature position based on the keywords; determines the signature rules based on the current review process; merges the corresponding signature nodes, signature positions, signature files, and signature rules and transmits them to the contract lock platform for signing; the adversarial detection model is based on a neural network structure, and its weights and parameters are determined through early adversarial training and later dynamic threshold adjustment, which is used to perform four-property detection on the electronic files; the electronic file storage module includes a general storage unit and a stream file storage unit, the general storage unit is used to store the original electronic files, and the stream file is used to store the file stream containing signature stream data; the archiving and handover module is used to archive and hand over the electronic files that pass the four-property detection.

[0014] In the present invention, preferably, a storage medium is also proposed, which stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the method for electronic file organization and single-set file transfer under the four-dimensional environment as described above.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] The method of the present invention discovers hidden logical errors by constructing a metadata topology map and a graph neural network, and adopts a multi-dimensional joint evaluation mechanism. It analyzes time series characteristics through an LSTM network and comprehensively judges the authenticity, integrity, availability and security of files in combination with cryptographic strength indicators; at the same time, it improves the approval efficiency of engineering electronic files and enhances the integration of electronic signature processes and project management systems; the coverage of the four-property detection is extended to defect types that are difficult to identify with traditional methods, such as time series anomalies and metadata logical errors; the integrity verification of archived files continuously optimizes the adversarial detection model through the dynamic generation of adversarial samples, which can effectively reduce the risks of missed detection and misjudgment; at the same time, the file transfer process adopts encrypted storage and verification matrix to ensure the readability and tamper-proofness of data during long-term storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of the method for electronic file organization and single-set file transfer under the four-property environment described in the present invention.

[0018] Figure 2 This is a flow chart of the adversarial training strategy for the electronic file organization and single-set file transfer method under the four-property environment described in the present invention.

[0019] Figure 3 This is a schematic diagram of the authenticity detection process in the method for electronic file organization and single-set file transfer under the four-quality environment described in the present invention.

[0020] Figure 4 This is a schematic diagram of the dynamic threshold adjustment process of the adversarial detection model in the method for electronic file organization and single-set file transfer under the four-property environment described in the present invention.

[0021] Figure 5 This is a structural diagram of the electronic file organization and single-set file transfer system under the four-property environment described in the present invention. DETAILED DESCRIPTION

[0022] The technical solutions of this application will be described clearly and completely below, in conjunction with the accompanying drawings. It should be understood that the described embodiments represent only a portion of the embodiments of this application, and not all of them. The components of this application, generally described and illustrated in the drawings herein, may be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but rather merely represents selected embodiments of this application. All other embodiments derived by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. Furthermore, in the description of this application, the terms "first," "second," etc., are used solely to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] Construction projects generate a massive amount of electronic documents, and their importance is growing. Traditional engineering document management relies on manual labor, which is inefficient and prone to errors. While electronic signature technology has been partially implemented, it suffers from issues such as insufficient platform integration, poor interface compatibility, and incomplete permissions management. Existing document verification methods often rely on single-dimensional detection techniques, making it difficult to simultaneously meet the comprehensive assessment requirements for authenticity, integrity, availability, and security. This is particularly inefficient when processing complex formats and large amounts of data.

[0024] Please also see Figures 1 to 4 A preferred embodiment of the present invention provides a method for electronic file organization and single-set file transfer under a four-property environment, including collecting electronic files during the construction period; adding the electronic files to the approval process and electronically signing them to generate a file stream containing signature stream data; pre-archiving and classifying the electronic files that have completed approval; using a detection strategy based on an adversarial detection model to perform four-property detection on the pre-archived files; registering the electronic files from the construction period that pass the detection into the warehouse for archiving and transfer.

[0025] Electronic documents during the construction phase refer to structured or unstructured data generated throughout the project lifecycle. These include pre-construction document submissions and commencement orders, as well as documents related to quality management, progress management, daily controls, safety management, and inspections during construction, as well as final acceptance inspections of completed defects and the defect liability period. These documents can be collected through IoT sensors, BIM systems, and project management software, forming a collection of electronic documents including design drawings, construction logs, and acceptance reports. Electronic documents can be uploaded en masse or in batches based on progress. Document data for the same project is consolidated into a single electronic document. Once complete, the electronic document enters the approval process. Upon approval, it can be electronically signed or directly transferred to the approved document. When electronic signatures are required during the approval process, the electronic signature process is implemented through integration with a third-party authentication platform. RSA encryption algorithms and a timestamp server can be used to generate tamper-proof signature stream data, forming a complete chain of evidence that includes the signatory's identity, time point, and approval status. After approval, electronic documents are pre-archived. Pre-archiving classification is accomplished by assigning project stage and file type tags. The pre-archived electronic documents are tested for four properties, namely authenticity, integrity, availability and security. An adversarial detection model is used for intelligent detection. The adversarial detection model is built based on a deep neural network. In the early stage, it is iteratively trained with training data containing forged features so that the adversarial detection model can identify four defects in electronic documents, such as abnormal signatures, tampering traces, certificate chain anomalies, timestamp conflicts, and missing metadata. Based on the results of the adversarial detection model, the four-property test can be completed quickly, while ensuring the confidential transfer of electronic documents and realizing efficient electronic document archiving. The entire processing process is safe and reliable.

[0026] Traditional methods rely on manual verification of signature authenticity or simple summary verification to determine signature authenticity. However, this application uses an adversarial detection model to automatically identify forged certificates and abnormal timestamps, making verification more accurate and efficient. Existing technologies often use fixed rules to detect file authenticity and integrity. This method discovers hidden logical errors by constructing a metadata topology map and graph neural network. Traditional file verification systems independently process four indicators. This solution adopts a multi-dimensional joint evaluation mechanism, analyzing time series characteristics through an LSTM network and combining cryptographic strength indicators to comprehensively judge file authenticity, integrity, availability, and security.

[0027] The above technical solution improves the efficiency of electronic engineering document approval and enhances the integration of electronic signature processes with project management systems. The coverage of the four-property detection system has been expanded to include defect types difficult to identify with traditional methods, such as time series anomalies and metadata logic errors. Archived file integrity verification uses dynamic adversarial sample generation to continuously optimize the adversarial detection model, effectively reducing the risk of missed detections and false positives. Furthermore, the file transfer process utilizes encrypted storage and a verification matrix to ensure data readability and tamper resistance during long-term storage.

[0028] In this embodiment, the adversarial detection model adopts an adversarial training strategy, including creating a virtual signature node to generate adversarial stream data containing forged certificate chains and abnormal timestamps, constructing metadata conflict samples, mixing the adversarial stream data with real data in a preset proportion to generate a training set, and performing multiple adversarial training until the misjudgment rate is lower than the threshold.

[0029] Among them, a virtual signature node refers to a fake process node generated by simulating an unauthorized signature operation in the approval process. It is generated using preset illegal user identity parameters and forged permission identifiers, and is used to simulate unauthorized file tampering scenarios. A forged certificate chain refers to a sequence of digital certificates issued by non-certification authorities or with substandard key lengths. It is generated by modifying the certificate authority field and shortening the RSA key length to less than 512 bits. It is used to test the adversarial detection model's ability to identify illegal authentication credentials. An abnormal timestamp refers to a time stamp that conflicts with the logical time window of the approval process. For example, a final review timestamp is generated before the approval process is started, or the timestamps of each node form a reverse sequence. This is used to detect timing logic anomalies. Metadata conflict samples are abnormal data instances that violate the metadata specifications of engineering files. They are constructed by randomly deleting required fields, filling integer fields with string content, and repeatedly generating metadata records with the same field name but different values. These metadata are generated as adversarial sample data and used to train adversarial detection models to identify data structure anomalies.

[0030] Specifically, the implementation process of the adversarial training strategy is divided into two stages: adversarial sample data construction and adversarial detection model iterative optimization. In the adversarial sample data construction stage, abnormal timestamps are constructed through virtual signature nodes and forged certificate chains to generate adversarial file streams. At the same time, metadata conflict samples containing missing fields, type errors, and repeated definitions are constructed. These adversarial samples cover the typical abnormal forms of electronic files in the aspects of certificate legitimacy verification, approval timing logic, and data structure integrity. In the adversarial detection model training stage, the constructed adversarial sample data and real engineering files are mixed in a preset ratio and input into the detection model. Through multiple training iterations, the model establishes an association mapping between abnormal features and error types. During the training process, the adversarial detection model's misjudgment rate indicator for the mixed data set is continuously monitored. When the adversarial detection model can stably identify adversarial samples and the misjudgment rate for real data is lower than the preset threshold, the training is terminated, thereby ensuring that the adversarial detection model has the ability to distinguish between real business data and potential attack samples.

[0031] Traditional engineering document detection methods typically use fixed rules to independently test the authenticity, integrity, availability, and security of electronic documents, resulting in low detection efficiency and accuracy. Alternatively, they rely on historical real-world data for intelligent model training, which can lead to insufficient generalization when faced with new attack vectors or anomalous data. This method proactively generates adversarial samples containing multi-dimensional attack signatures and employs a dynamic mixed-ratio training strategy. This enables the adversarial detection model to comprehensively cover anomalous patterns that may arise during the flow of engineering documents, enabling it to identify the authenticity, integrity, availability, and security of electronic documents and improving the accuracy of identifying unknown error types.

[0032] Through the above technical solutions, this application effectively solves the problems of low efficiency, poor accuracy and insufficient generalization ability of traditional detection methods. By actively constructing training data with attack characteristics, the detection model can accurately identify malicious files containing forged certificate chains, abnormal timestamps and metadata conflicts, and control the error rate of the four-property detection of engineering files within an acceptable range. The adoption of a hybrid training strategy not only retains the adversarial detection model's ability to process normal business data, but also enhances its defense performance against new attack methods, achieving a dual improvement in the robustness and accuracy of the detection model.

[0033] In this embodiment, specific implementation steps for electronic archive authenticity detection based on the four-property environment are proposed: extracting the timestamp sequence in the file stream and constructing the time series feature vector through the LSTM network; comparing the actual signature node with the time window matching of the approval process; calculating the cryptographic strength index of the certificate chain, including key length, hash collision probability and signature timeliness; inputting the time series feature vector, time window matching and cryptographic strength index into the adversarial detection model after adversarial training, and outputting the credibility score.

[0034] The timestamp sequence refers to a time-stamped data set generated at each operation node in the approval process for electronic documents. Timestamp fields containing creation, modification, and approval times are stored in JSON format to reflect the dynamic temporal characteristics of the document flow. An LSTM network, a long short-term memory neural network, employs a bidirectional network structure with 64 hidden units in this implementation. A sliding window is used to intercept timestamp sequence segments to capture unusual fluctuations in time intervals within the approval process. Time window matching measures the degree of conformity between the actual signature operation time and the preset time range of the corresponding node in the approval process. The matching score is calculated using the absolute value of the time difference between the timestamp and the preset time range of the process node, and is used to detect illegal operations or process skipping. The certificate chain cryptographic strength index comprises three dimensions: key length can be set to a parameter of 2048 bits or more in the RSA algorithm; hash collision probability can be assessed by calculating the theoretical collision value of the SHA-256 algorithm; and signature validity is determined by verifying the overlap between the digital certificate's validity period and the file's creation time, forming a multi-dimensional security assessment system.

[0035] Specifically, in the electronic document approval process for engineering projects, the timestamp data generated at each operation node is extracted as a continuous time series. This sequence is modeled using an LSTM network to capture the temporal distribution characteristics of different approval stages, such as the time interval distribution pattern in a normal process. When abnormal approval operations occur, such as when the same user operates between different nodes with extremely short intervals or inverse time order, the time series feature vector output by the LSTM will fluctuate significantly. Furthermore, a preset time window for the approval process defines the acceptable range of operation times for each node. By calculating the deviation between the actual signature time and the window boundary, operations that violate the process's timing constraints can be identified. Regarding cryptographic verification, a comprehensive assessment is conducted to ensure that the key length of the certificate chain meets security standards, the hash algorithm is collision-resistant, and the digital signature is within its validity period, thereby avoiding single-dimensional security verification vulnerabilities. These multi-dimensional features are normalized and then fed into an adversarial detection model. The neural network model learns the nonlinear correlations between the features and ultimately outputs a comprehensive credibility score. When the score falls below a preset threshold, a manual review process is automatically triggered.

[0036] Traditional authenticity detection methods typically rely solely on static timestamp verification and key length validation. For example, existing technologies may only check the legitimacy of a single timestamp, failing to capture the dynamic correlation between multiple timestamps within the approval process. Certificate verification may also focus solely on key length compliance, ignoring hash algorithm strength and signature timeliness. This solution enables dynamic monitoring of the approval process through temporal feature modeling and constructs a composite security assessment mechanism using multi-dimensional cryptographic indicators, addressing the single-dimensional nature of traditional methods.

[0037] Through the above technical solution, this application realizes dynamic time-series monitoring of the electronic document approval process, effectively identifying abnormal fluctuations in the timestamp sequence; establishes a time-series correlation verification mechanism between signature nodes and approval processes to prevent illegal operations such as process skipping or time inversion; and constructs a multi-dimensional cryptographic strength assessment system to avoid verification vulnerabilities caused by defects in a single security dimension.

[0038] In this embodiment, a detection strategy for integrity detection steps is proposed, which specifically include: constructing a metadata topology graph to analyze the logical dependencies between fields; detecting abnormal connection edges and missing nodes through a graph neural network; calculating the semantic similarity between the file entity and the metadata description; and dynamically generating adversarial text containing randomly missing fields to verify the misjudgment rate of the adversarial detection model.

[0039] Among them, the metadata topology graph refers to a network structure diagram that visually presents the logical relationships between electronic file metadata fields. This can be achieved, for example, by using graph database technology to construct an entity-attribute-relationship model. This feature can structurally represent the nested relationships of multi-level metadata in complex engineering files. Among them, the graph neural network refers to a deep learning model that performs feature learning based on graph-structured data. For example, it can be achieved by using a graph attention network or a graph convolutional network. This feature effectively identifies abnormal connection patterns by automatically extracting node features and edge weights from the metadata topology graph. Among them, semantic similarity calculation refers to an indicator that quantifies the degree of consistency between file content and metadata descriptions. For example, it can be achieved by using a pre-trained language model to extract text vectors and then calculate cosine similarity. This feature implements a cross-validation mechanism for the content layer and description layer. Dynamic generation of adversarial text refers to the automatic construction of test samples with specific defects through an algorithm. For example, it can be achieved by using Markov chain random sampling combined with field masking technology. This feature continuously verifies the robustness of the model by actively injecting abnormal data.

[0040] Specifically, when processing engineering files containing complex metadata structures, a topology graph is first constructed based on the attributes and constraints of the metadata fields, with each metadata field as a node and the logical dependencies between fields as edges. The topology graph is then traversed using a graph neural network to identify abnormal connection edges that violate business rules and missing nodes with undefined data types. Simultaneously, natural language processing techniques are used to semantically parse the file body content and calculate a similarity threshold with the metadata description. To further verify the reliability of the detection model, the system automatically generates test text containing randomly missing fields. By observing the model's recognition accuracy for abnormal samples, the sensitivity parameters of the detection strategy are dynamically adjusted.

[0041] Compared to existing technologies, traditional integrity checks primarily rely on pre-set rule engines for format verification, which cannot effectively handle metadata fields with dynamic relationships. Conventional methods typically use static rule base matching to detect metadata topology anomalies, which is difficult to adapt to the ever-changing field association patterns in engineering files. This solution, however, uses a graph neural network to automatically learn the underlying association patterns of metadata fields. Combined with dynamically generated adversarial examples, it continuously optimizes the model, significantly improving anomaly detection coverage in complex scenarios.

[0042] Through the above technical solution, this application solves the technical problems of inaccurate identification of abnormal connections and lack of semantic consistency verification in engineering file integrity detection. By combining the metadata topology map with the graph neural network, a deep analysis of the relationship between complex fields is achieved; a two-way verification mechanism for content and description is established through semantic similarity calculation; and dynamic adversarial text generation technology ensures that the detection model can continuously adapt to new data anomaly patterns and form a self-optimizing detection closed loop. In a specific embodiment, this method and the traditional method are used to detect the electronic archives of the same completed project. The detection comparison results are shown in Table 1 below. The original method can construct a metadata topology map with 5304 nodes and 7821 edges in a short time, detect 12 abnormal associations (such as illegal mapping of concrete grade and seismic grade), and improve the recognition rate of field missing attacks from 70% to 93% through the adversarial detection model.

[0043] Table 1.

[0044]

[0045] In this embodiment, a detection strategy for availability detection steps is proposed, specifically including: determining whether the format of the pre-archived electronic file meets the requirements of long-term storage, determining whether the metadata of the pre-archived electronic file is accessible and whether the metadata in the information package is readable.

[0046] Among them, the requirement for long-term storage means that the file format has backward compatibility and cross-platform parsing capabilities. In this embodiment, the long-term preservation format standard defined by the International Organization for Standardization is adopted to achieve this. For example, electronic files are converted to PDF / A format to ensure that the files can still be correctly parsed by standard software decades later. Metadata accessibility means that the system can obtain metadata through an API interface or database query method. For example, the OAuth2.0 authorization protocol is used to implement access permission control to ensure that authorized users can retrieve metadata fields in real time. Metadata readability in the information package means that the metadata maintains a physical storage association with the file entity it describes. The binding storage of metadata and file entities is achieved through the XML encapsulation specification, and the data readability is verified by verifying the integrity of the encapsulation structure.

[0047] Specifically, during the pre-archiving stage of electronic documents, an availability assurance framework is established through a dual verification mechanism. First, the file format is technically screened, and only open format standard files are retained, eliminating the risk of unreadable files due to accidental deletion at the carrier level. At the same time, layered detection is implemented on the metadata, verifying whether the permission configuration of the access interface complies with the security policy at the logical level, and detecting the structural integrity of the metadata package at the physical level. For example, when a disconnect is detected between the metadata storage path and the file entity, an automatic repair mechanism can be triggered to re-establish the index association. This dual verification from format to metadata can detect potential availability issues caused by packaging errors or improper permission configuration.

[0048] Compared to existing technologies, traditional availability checks typically only verify file format compatibility, failing to establish a dual-check mechanism for metadata access permissions and physical storage. Existing methods are unable to detect metadata access restrictions caused by misconfigured permissions, nor can they identify broken links between metadata within an encapsulated package and the file entity. By adding a metadata accessibility verification step, this solution proactively identifies authorization policy vulnerabilities and, by verifying metadata readability within the information package, promptly remediates storage structure anomalies.

[0049] Through the above technical solution, this application effectively solves the problem of insufficient long-term readability of electronic files, avoiding the risk of future parsing failures due to the use of proprietary formats. It also overcomes the mismatch between metadata access rights and physical storage status, ensuring that archived electronic files are fully accessible within authorized scopes and that metadata and file entities always maintain an accurate correspondence.

[0050] In this embodiment, security detection steps are proposed, specifically including: generating attack simulation data of data tampering attacks, signature forgery attacks and structure destruction attacks, mixing the attack simulation data with real data in a preset ratio and putting them into the training set of the adversarial detection model for training until the adversarial detection model's misjudgment rate of the input data is lower than the threshold; using the adversarial detection model generated by gradient penalty optimization, and using the optimized adversarial detection model to perform security detection on electronic files.

[0051] Among them, data tampering attacks refer to the behavior of destroying data integrity by modifying file content or metadata, such as generating attack samples by random bit flipping, injecting malicious code, or tampering with hash values, which are used to train models to identify unauthorized modification behaviors. Signature forgery attacks refer to the behavior of counterfeiting legitimate electronic signatures or certificate chains, such as constructing adversarial samples by generating forged certificate chains, timestamp offsets, or reverse engineering signature algorithms, which are used to enhance the model's ability to identify forged signatures. Structural destruction attacks refer to the behavior of invalidating files by deleting key fields or disrupting the logical structure of files, such as generating training samples by randomly deleting metadata fields, disrupting the file encoding order, or inserting illegal characters, which are used to improve the model's recognition accuracy for abnormal structures. Gradient penalty optimization refers to imposing constraints on the gradients of parameter updates during model training, such as using regularization terms to limit the gradient amplitude or forcing a smooth transition in the parameter space, to prevent the model from overfitting specific attack patterns during adversarial training.

[0052] Specifically, the attack simulation data generation module simultaneously constructs three types of attack samples: data tampering, signature forgery, and structural destruction, covering the main types of attacks that electronic files may suffer. The attack samples and real data are mixed in a preset ratio and then input into the adversarial detection model for iterative training. During the training process, the model's misjudgment rate of normal files is continuously monitored. When the misjudgment rate is lower than the preset threshold, the training is terminated to ensure that the model controls the risk of false alarms while maintaining a high detection rate. The model that has completed preliminary training uses the gradient penalty optimization method to adjust its parameters. A gradient constraint term is introduced in the loss function to force the model parameters to maintain a smooth transition during the optimization process to avoid overfitting to specific attack patterns. The resulting optimized model can effectively distinguish normal files from abnormal files in mixed attack scenarios, while maintaining detection stability and improving the adaptability to new attacks.

[0053] Compared to existing technologies, traditional methods typically use samples from a single attack type for training, resulting in a significant decrease in detection performance when facing mixed attacks. Existing adversarial training processes lack dynamic optimization mechanisms, making them prone to overfitting to specific attack patterns and reducing generalization capabilities. Existing gradient optimization methods fail to consider the perturbation characteristics of adversarial samples, making it difficult to effectively balance detection accuracy and false positive rate.

[0054] Through the above technical solutions, this application effectively addresses the issue of insufficient detection coverage in hybrid attack scenarios. By combining training with multiple attack samples, the model's ability to identify complex attacks is enhanced. A dynamic false positive rate threshold control mechanism is employed to ensure detection accuracy while reducing the false interception rate of legitimate files. The introduction of a gradient penalty optimization method significantly improves the model's generalization performance, enabling it to adapt to evolving attack vectors.

[0055] This application further proposes a dynamic threshold adjustment method for the adversarial detection model, establishes a Bayesian probability distribution model of four indicators based on historical detection data, calculates the KL divergence of the current batch of detection results and the probability distribution in real time, and triggers the adversarial data generation module to update the training set when the divergence value exceeds the warning threshold, and adjusts the weight coefficient of each detection dimension to form closed-loop feedback and dynamic optimization.

[0056] Among them, the Bayesian probability distribution model refers to a probability distribution model of the four-property indicators based on historical detection data, for example, it is implemented by the kernel density estimation method, and is used to dynamically model the data distribution of the four-property detection results. KL divergence calculation refers to calculating the difference between the current batch detection results and the historical distribution, for example, it is implemented by the information entropy difference quantization method, and is used to detect the degree of data distribution deviation. The adversarial sample data generation module updates the training set and automatically generates new adversarial samples when data deviation is detected, for example, it is implemented by generative adversarial network technology, and is used to expand the diversity of training data. Weight coefficient adjustment refers to the reallocation of decision weights of the four-property detection dimensions according to the direction of data deviation, for example, it is implemented by the gradient backpropagation algorithm, and is used to balance the sensitivity of the adversarial detection model to different indicators.

[0057] Specifically, the Bayesian probability distribution model converts the historical four-property detection results into a probability distribution through the kernel density estimation method to form a dynamic benchmark. The KL divergence calculation module continuously monitors the difference between the current detection result and the benchmark distribution. When the difference exceeds the preset warning threshold, it triggers the adversarial data generation module to generate adversarial samples containing new attack patterns. The new samples are mixed with the original data to update the training set and retrain the adversarial detection model. At the same time, the gradient backpropagation algorithm dynamically adjusts the weight coefficients of each detection dimension based on the performance of the model on the new data. For example, when an increase in timestamp forgery attacks is detected, the weight ratio of the time window matching is increased. This forms a closed-loop feedback mechanism from data monitoring to model updating, so that the detection threshold is automatically optimized as the data distribution changes.

[0058] Compared to existing technologies, traditional methods using fixed thresholds are unable to adapt to shifts in the four-property detection data caused by changes in project file types or new attack methods. For example, existing static threshold models may misjudge unknown signature forgery attacks because the detection rules do not cover the new attack characteristics. This solution, through dynamic probabilistic modeling and KL divergence quantification, can perceive changes in data distribution in real time. It utilizes an adversarial sample generation mechanism to actively expand the model's detection boundaries, and combines weight adjustment to optimize the decision balance between detection dimensions, effectively addressing detection bias caused by dynamic data evolution.

[0059] Through the above technical solution, this application solves the problem of increased false positives in static threshold models due to data distribution shifts. Through dynamic probability modeling and a closed-loop feedback mechanism, the detection model can automatically adapt to scenarios such as changes in engineering file types and the introduction of new attack methods, thereby improving the accuracy and robustness of the four-property detection. For example, when a new electronic signature format is added to the engineering file management system, the model can automatically identify changes in the signature timestamp distribution, generate corresponding adversarial samples to update the training set, and adjust the weight parameters of the time window matching degree to maintain detection accuracy.

[0060] The present application further proposes that the electronic file package to be archived and transferred includes the encrypted original electronic file, the complete file stream with time stamp, and the verification matrix.

[0061] Among them, the encrypted original electronic file refers to the electronic file content encrypted through a cryptographic algorithm, such as the AES-256 encryption algorithm combined with digital envelope technology, and access control is achieved through an encryption key management system. This feature is used to prevent files from being illegally stolen or tampered with during transmission or storage, ensuring data confidentiality. A complete file stream with a timestamp refers to the time-series record data of all operation nodes of an electronic file, such as the use of blockchain technology to generate an unalterable timestamp sequence, which is calibrated with a time synchronization server and a standard time source. This feature is used to establish a traceable chain of operational evidence and provide time dimension support for authenticity verification. The verification matrix refers to a data structure composed of file metadata, hash values ​​and verification rules, such as the use of sparse matrix coding technology, and a field association relationship map generated by a preset verification algorithm. This feature is used to quickly locate abnormal associations between file content and metadata, thereby improving the efficiency of integrity verification.

[0062] Specifically, encrypted original electronic files are stored on different media using a key splitting mechanism during archiving, with data only restored through multi-party collaborative decryption during the transfer process. The timestamped file stream uses a timestamp server to solidify the operation time of each signature node into the blockchain distributed ledger, forming a time-series evidence chain encompassing the entire approval process. The verification matrix extracts key fields from the file metadata to generate a hash fingerprint, constructs a matrix index based on logical dependencies, and during the verification phase, compares the mapping relationship between hash values ​​and metadata through matrix operations to quickly locate abnormal fields.

[0063] Compared with existing technologies, traditional electronic file packages contain only the original file and simple metadata, lack cryptographic protection measures and time dimension verification mechanisms, and integrity verification requires item-by-item hash value comparison. This solution builds a three-in-one security verification system through structured encapsulation of encrypted files, timestamped file streams, and a verification matrix, creating synergistic effects in three dimensions: data encryption and storage, operation timing fixation, and verification efficiency optimization.

[0064] Through the above technical solutions, this application strengthens the tamper-proof capabilities of electronic files during transmission, reduces the time cost of authenticity verification through a timestamp evidence chain, and uses a matrix data structure to reduce the computational complexity of integrity verification from linear to logarithmic levels. The combination of encryption mechanism and verification matrix effectively blocks unauthorized access paths, and the timestamp file stream provides a verifiable time reference point for audit tracing.

[0065] See Figure 5 , this application further proposes an electronic archive organization and single-set archive transfer system under a four-property environment, including a receiving module, an approval module, an electronic signature module, an adversarial detection model, an electronic file storage module, and an archiving and handover module. The receiving module is used to receive electronic files during the construction period; the approval module initiates an approval process for the electronic files and conducts step-by-step approval; the electronic signature module communicates with the contract lock platform to electronically sign the electronic files: extract the keywords of the electronic files that need to be signed, and determine the signature position based on the keywords; determine the signature rules based on the current review process; merge the corresponding signature nodes, signature positions, signature files and signature rules and transmit them to the contract lock platform for signing; the adversarial detection model is based on a neural network structure, and its weights and parameters are determined through early adversarial training and later dynamic threshold adjustment, which is used to perform four-property detection on electronic files; the electronic file storage module includes a general storage unit and a stream file storage unit, the general storage unit is used to store the original electronic files, and the stream file is used to store the file stream containing signature stream data; the archiving and handover module is used to archive and hand over the electronic files that pass the four-property detection.

[0066] The receiving module is used to receive data from the construction period, which includes the entire process from before construction to during construction and after completion. The data before construction includes the certification of the construction / supervision / agent project department, personnel certification, five-party application materials and all approved data; the data for construction process control includes all data of progress control, quality management, safety management and quality and safety inspection processes; the data after completion includes all data of completion management and defect liability period. The construction period data received by the receiving module are sorted and merged according to the project number and cycle. When all the data for the same construction period are complete, they are sent to the approval module, which initiates the approval process. The approval process is carried out in sequence according to the set rules. When the approval is passed, the electronic signature module is activated for electronic signature in the process that requires electronic signature. The process that does not require signature will automatically flow to the next approval process until all approvals are completed.

[0067] Specifically, after receiving the electronic file, the receiving module first classifies the electronic file, including structured data and unstructured data, performs data cleaning on the structured data, and pre-processes the unstructured data. The processed data is uniformly verified, the project number is coded, and the signed data is stored in the storage module in cycles. When the entire project cycle is completed, it enters the approval module. The approval module is mainly based on the intelligent approval engine and is carried out through a preset approval process. When each approval process is passed, it determines whether a signature is required. When a signature is required, the electronic signature module is called, otherwise it is automatically filed according to the project number. The location recognition of the electronic signature can be determined based on OCR combined with keyword matching positioning.

[0068] Among them, the electronic signature module communication connection contract lock platform refers to the data intercommunication between the system and the third-party electronic signature platform through the standard interface protocol, such as the use of an integration method based on a RESTful API to realize the automatic transmission of signature requests and the return of results. The adversarial detection model based on the neural network structure refers to the use of a hybrid architecture that combines a multi-layer convolutional network with a graph neural network. The adversarial detection model can be trained through the TensorFlow framework to enable it to process structured metadata and unstructured file content at the same time. Dynamic threshold adjustment refers to the use of a Bayesian probability model to model the distribution of detection results, such as using KL divergence to calculate the deviation between real-time data and historical distribution to trigger model parameter updates. The stream file storage unit is used to store file streams containing signature stream data, which means saving the timestamp, node sequence and certificate chain information of the signature operation in the form of incremental logs.

[0069] Specifically, the electronic signature module uses a keyword extraction algorithm to identify key clauses requiring signatures within electronic documents. For example, regular expression-based semantic matching technology is used to locate fields such as the contract amount and the name of the signatory. The signature rule generation module automatically matches a pre-set permission matrix based on the order of the current approval process. For example, within a three-level approval process, the electronic signature nodes for the department head, legal affairs officer, and project director are sequentially triggered. During the training phase, the adversarial detection model injects adversarial samples such as forged certificate chains and anomalous timestamps to enable the model to learn to distinguish between authentic signature data and attack data. For example, 20% of anomalous samples are mixed into the training set to improve model robustness. The stream file storage unit, while storing the original electronic document, records the time series data of all signature operations. For example, the action type, performer identity, and timestamp of each signature are appended to the file stream in JSON format. Upon successful detection, the archiving and transfer module packages the encrypted original document, signature stream data, and verification matrix into a standardized file package. For example, the file is encrypted using the AES-256 algorithm and a verification matrix containing a hash checksum is generated.

[0070] Compared with existing technologies, traditional engineering management systems rely on manual configuration of signature rules, which leads to permission conflicts and process breaks during cross-platform integration. This solution realizes automated matching of signature processes by dynamically generating signature rules and deeply connecting with the contract lock platform. Existing detection methods mostly use static rule bases for four-property verification, which cannot identify new attack methods. This solution uses adversarial training to enable the model to continuously adapt to changes in attack patterns, while using streaming storage technology to retain complete signature process data, providing traceability support for post-audit. Existing file storage solutions usually store metadata separately from physical files, making integrity verification difficult. This solution uses a dual storage unit structure to ensure the relevance and consistency of original files and signature stream data.

[0071] Through the above technical solution, this application realizes the integrated integration of engineering electronic document signing process and management system, solves the problem of inconsistent signing authority across platforms through keyword positioning and dynamic matching of rules; uses adversarial training neural network model to improve the detection accuracy of four abnormal features, and ensures the complete traceability of file data throughout its life cycle through streaming storage structure; standardized archiving process ensures that the transferred file package meets compliance requirements and avoids omissions or format errors caused by manual operations.

[0072] In some other preferred embodiments of the present invention, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the method described in the above embodiment.

[0073] Among them, storage medium refers to the physical carrier that stores computer programs, such as a solid-state drive, disk array or cloud storage server, to ensure that program instructions can be called and executed by the processor.

[0074] Among them, a computer program refers to a set of instructions containing data processing logic, such as one written in an object-oriented programming language, to achieve automated control of the electronic document signing process and the call of the four-property detection algorithm.

[0075] Among them, processor execution refers to the calculation and processing of program instructions through the central processing unit or graphics processing unit, such as using a multi-threaded parallel computing architecture to simultaneously process the signature verification and four-property detection tasks of electronic documents.

[0076] Specifically, when the computer program stored in the storage medium is loaded and executed, it establishes a communication connection with the contract lock platform through the electronic signature module, automatically matches the signature location based on file keyword extraction technology, and dynamically generates signature rules based on the review process, realizing deep integration between the engineering management system and the signature platform. The adversarial detection model uses a neural network structure to conduct adversarial training on injected forged certificate chains, abnormal timestamps, and metadata conflict samples, establishes a dynamic threshold adjustment mechanism, and optimizes the detection weight coefficient in real time to respond to new attack modes. The streaming file storage unit works in conjunction with the ordinary storage unit to record the original electronic file and the signature process data containing the timestamp sequence, forming a complete chain of evidence. The archiving and transfer module encrypts and encapsulates the electronic files that pass the inspection, generating a standardized data package containing a verification matrix to ensure the long-term verifiability of the transferred files.

[0077] Compared with existing technologies, existing storage media mostly use a single program to execute fixed detection rules, which cannot adapt to compatibility issues caused by differences in the interfaces between engineering management systems and signature platforms. This solution solves the problems of signature position dislocation and process interruption caused by inconsistent interface protocols in traditional systems through dynamic signature rule matching and a dual-track storage structure for streaming files. Existing adversarial detection models mostly rely on static sample library training, while this solution constructs a dynamic training set by injecting virtual signature nodes and metadata conflict samples, giving the model a continuously evolving anomaly detection capability. The existing archiving and transfer process lacks a mechanism to associate timestamp file streams with verification matrices. This solution improves the traceability of archives after transfer through a composite encapsulation method that encrypts the original file and generates a verification matrix.

[0078] Through the above technical solution, this application realizes the seamless connection between the engineering management system and the electronic signature platform, eliminating the problem of signature process interruption caused by interface protocol differences; through adversarial training and dynamic threshold adjustment mechanism, the four-property detection model can identify new forgery attacks and format tampering behaviors in real time; through the dual-track storage mode of streaming files and ordinary files, the full life cycle data of electronic files from generation to archiving is fully recorded; through the encryption packaging method including timestamp file stream and verification matrix, the verifiability and tamper-proof capability of the transferred archives during long-term storage are ensured.

[0079] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0080] The above description is a detailed description of the preferred embodiments of the present invention, but the embodiments are not intended to limit the scope of the patent application of the present invention. Any equivalent changes or modifications completed under the technical spirit suggested by the present invention should fall within the patent scope covered by the present invention.

Claims

1. The method for organizing electronic files and transferring single-set files under the four-factor environment is characterized by: include: Collect electronic documents during the construction period; Add electronic documents to the approval process and perform electronic signatures to generate a file flow containing signature flow data; Pre-archive and classify the electronic documents that have completed approval; Adopting a detection strategy based on an adversarial detection model to perform four-property detection on pre-archived files, wherein the four properties include authenticity, integrity, availability, and security; After passing the inspection, the electronic documents of the construction period will be registered and stored in the warehouse for archiving and handover; The adversarial detection model adopts an adversarial training strategy: Create a virtual signing node and generate adversarial stream data containing forged certificate chains and abnormal timestamps; Construct metadata conflict samples, including abnormal sample data with enumeration value out of bounds, format misplacement, and repeated fields; Mixing adversarial flow data, abnormal sample data, and real data in a preset ratio to generate a training set, and using the training set to perform adversarial training on the adversarial detection model multiple times until the adversarial detection model's misjudgment rate for input data is lower than a threshold; Also includes dynamic threshold adjustment for adversarial detection models: Establish a Bayesian probability distribution model of the four indicators based on historical test data; Calculate the KL divergence between the test results of the current batch and the probability distribution in real time; When the divergence value exceeds the warning threshold, the adversarial data generation module is triggered to update the training set; Adjusting the weight coefficients of each detection dimension to form closed-loop feedback and dynamic optimization of the adversarial detection model; The detection strategy for the authenticity detection specifically includes: Extract the timestamp sequence from the file stream and construct the time series feature vector through the LSTM network; Compare the actual signature node with the time window of the approval process; Calculate cryptographic strength metrics for the certificate chain, including key length, hash collision probability, and signature timeliness; Inputting the time series feature vector, time window matching degree and cryptographic strength index into the adversarial detection model after adversarial training, and outputting a credibility score; The detection strategy for the integrity detection specifically includes: Build metadata topology diagrams and analyze logical dependencies between fields; Detect abnormal connection edges and missing nodes through graph neural networks; Calculate the semantic similarity between the file entity and the metadata description; Dynamically generate adversarial text containing randomly missing fields to verify the false positive rate of the adversarial detection model; The detection strategy for the security detection specifically includes: Generate attack simulation data for data tampering attacks, signature forgery attacks, and structural destruction attacks, mix the attack simulation data with real data in a preset ratio and put them into the training set of the adversarial detection model, and train the adversarial detection model until the adversarial detection model's misjudgment rate for input data is lower than a threshold; An adversarial detection model generated by gradient penalty optimization is used, and the optimized adversarial detection model is used to perform security detection on electronic files.

2. The method for organizing electronic files and transferring single-set files under a four-factor environment according to claim 1 is characterized in that: The detection strategy for the availability detection specifically includes: Determine whether the format of pre-archived electronic files meets the requirements for long-term storage; Determine whether the metadata of pre-archived electronic files is accessible and whether the metadata in the information package is readable.

3. The method for electronic file organization and single-set file transfer under a four-factor environment according to claim 1 is characterized in that: The electronic file package to be archived and transferred includes the encrypted original electronic file, the complete file stream with time stamp, and the verification matrix.

4. A system for organizing electronic files and transferring single-set files under a four-dimensional environment, for implementing the method for organizing electronic files and transferring single-set files under a four-dimensional environment as described in any one of claims 1 to 3, characterized in that: Including receiving module, approval module, electronic signature module, adversarial detection model, electronic file storage module, archiving and handover module, The receiving module is used to receive electronic files during the construction period; The approval module initiates an approval process for the electronic document and performs approval step by step; The electronic signature module is connected to the contract lock platform to electronically sign the approved electronic documents; Extract keywords from electronic documents that need to be signed, and determine the signature location based on the keywords; Determine the signing rules based on the current review process; The corresponding signature node, signature location, signature file and signature rules are combined and transmitted to the contract lock platform for signing; The adversarial detection model is based on a neural network structure, and its weights and parameters are determined through early adversarial training and later dynamic threshold adjustment, so as to perform four-property detection on the electronic file; The electronic file storage module includes a common storage unit and a stream file storage unit, wherein the common storage unit is used to store the original electronic file, and the stream file storage unit is used to store the file stream containing the signature stream data; The filing and handover module is used to file and hand over electronic files that have passed the four-property test.

Images