An intelligent management method and system for bidding information based on artificial intelligence

Abstract

The present invention discloses an intelligent management method and system for bidding information based on artificial intelligence, which belongs to the field of information security technology, and specifically comprises: collecting hardware fingerprints and network behavior patterns of user operation terminals to generate composite biometric feature codes; extracting the document structure of bidding documents, generating fingerprint maps through hash algorithms, and binding and storing them with composite biometric feature codes; dynamically selecting chaotic encryption algorithm parameters to encrypt files according to network delay and packet loss rate, dividing them into variable-length data blocks and setting dynamic check codes, and randomly distributing encrypted files to discrete storage nodes; verifying the terminal features of the requesting party in real time, and triggering secondary verification if they do not match; reconstructing files after verification, obtaining data blocks from low-latency nodes, cross-verifying abnormal blocks and replacing them, and updating check codes; analyzing user file operation behaviors, identifying abnormalities and storing the results in a defense strategy library.

Description

Technical Field

[0001] The present invention relates to the field of information security technology, and in particular to an intelligent management method and system for bidding information based on artificial intelligence. Background Art

[0002] Bidding is a competitive transaction method widely used in many fields such as business and engineering. It aims to select the most advantageous partners through fair, just and open processes. With the advancement of the digital age, bidding activities are increasingly relying on information technology to improve efficiency and expand the scope of participation. The rise of online bidding platforms has enabled a large amount of bidding information to be quickly transmitted and processed, greatly changing the traditional bidding model, but the information management problems behind it have also become more prominent.

[0003] At present, bidding information management technology mainly relies on conventional encryption methods and simple user authentication methods. In the process of data transmission and storage, fixed encryption algorithms are generally used, and encryption strategies cannot be flexibly adjusted according to the real-time network conditions. In terms of user authentication, it is mostly a combination of user name and password, or supplemented by a simple SMS verification code, which makes it difficult to accurately identify the true identity of the user and abnormal operation behavior. For massive bidding documents, the storage method is relatively simple, and there is a lack of in-depth correlation analysis between file features and user operation features, resulting in cumbersome data tracing and verification processes and poor accuracy.

[0004] There are many urgent problems to be solved in the bidding document management process. On the one hand, during the transmission and storage of files, they are very vulnerable to the risk of data theft, tampering or loss due to factors such as network attacks and storage device failures. Once the traditional encryption method is cracked, the file content will be completely exposed. On the other hand, in the analysis of file operation behavior, the existing technology can only perform simple operation statistics and it is difficult to effectively identify potential abnormal operation behaviors, which poses a serious threat to the fairness and security of bidding activities. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent bidding information management method and system based on artificial intelligence to solve the following technical problems:

[0006] Existing technologies can only perform simple operation statistics and are unable to effectively identify potential abnormal operation behaviors.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] An intelligent management method for bidding information based on artificial intelligence, comprising the following steps:

[0009] S1. During the bidding data upload stage, the hardware fingerprint and network behavior pattern of the user's operating terminal are collected to generate a composite biometric code including keystroke rhythm, cursor movement trajectory and file access habits;

[0010] S2. Extract features from the uploaded bidding documents, identify document structure features, metadata timestamp sequence and version change pattern, generate document fingerprint map through hash algorithm, and bind and store the document fingerprint map with the composite biometric code;

[0011] S3. Dynamically select chaotic encryption algorithm parameters to generate encrypted bidding documents based on current network delay and packet loss rate, divide the encrypted bidding documents into data blocks of variable length, and dynamically check each data block; randomly distribute the encrypted data blocks to geographically discrete storage nodes, and each storage node only stores discontinuous data fragments;

[0012] S4. During the data access phase, the requester’s terminal hardware fingerprint and network behavior pattern are verified in real time. When a mismatch with the composite biometric code is detected, secondary verification is triggered;

[0013] S5. When the requester passes the verification, the bidding document is reconstructed, and data blocks are obtained in parallel from the three nodes with the lowest delay first, the continuity of the check code time series is verified, multi-node cross-verification is initiated for abnormal data blocks and automatically replaced, and the dynamic check codes of all data blocks are updated after the reconstruction is completed, and the details of the replacement operation are recorded;

[0014] S6. Analyze the user's behavior characteristics from accessing files to operating files, identify whether there are abnormal file operations, and store the detection results in the defense strategy library.

[0015] As a further solution of the present invention: In S1, the specific process of generating the composite biometric code includes:

[0016] The time interval between mouse clicks of three consecutive file upload operations by the user is recorded, the standard deviation is calculated and normalized into a behavioral feature vector; the correlation characteristics between the frequency of use of the backspace key and the cursor movement speed during document editing are extracted to generate an operation habit matrix; the waveform pattern of memory usage changes during multi-task switching is analyzed to extract the peak frequency and attenuation coefficient of the memory; the behavioral feature vector, operation habit matrix, waveform pattern and device MAC address are hashed and fused to generate a unique identification code containing a timestamp; the identification code is compared with the user's historical behavior database. If the deviation exceeds the preset threshold, an abnormal behavior warning is triggered.

[0017] As a further solution of the present invention: In S3, the chaotic encryption algorithm parameters are dynamically selected:

[0018] Monitor network delay jitter and packet loss ratio, calculate channel quality index, dynamically adjust parameter value of Lorentz equation according to channel quality index, generate chaotic sequence, encrypt original file data through chaotic sequence, set encrypted data block length to golden ratio of current network throughput; generate a chain check code for each data block, the chain check code includes hash value of previous data block, and embed a timestamp into the chain check code of each data block.

[0019] As a further solution of the present invention: in S3, randomly allocating the encrypted data blocks to storage nodes with discrete geographical locations includes:

[0020] Each storage node sets a different access protocol stack, inserts padding bytes of random length between adjacent data blocks, encrypts the index table that records the location of data blocks, and stores it in an independent geographical isolation area to ensure that the index information is separated from the data; the actual location of the storage node and the mapping relationship of the storage node in the system are rotated regularly, the status of each storage node is monitored in real time, and the data distribution is adjusted dynamically.

[0021] As a further solution of the present invention: in S4, the secondary verification includes:

[0022] Construct a temporal convolutional network structure, which includes multiple one-dimensional convolutional layers, pooling layers, and fully connected layers. The historical keyboard tapping sequence library and the historical mouse behavior data are spliced ​​into a multimodal behavior data tensor according to the time dimension and input into the temporal convolutional network. In the network training phase, the historical multimodal behavior data tensor is used as the training set, and the stochastic gradient descent optimization algorithm is used to minimize the cross entropy loss function, iteratively update the network weight parameters, and train to obtain the temporal correlation feature pattern of the multimodal behavior data;

[0023] Through the input device driver interface, keyboard tapping data is synchronously collected to obtain the original data sequence of tapping force and time interval; using the mouse device API, mouse behavior data is collected, including mouse click coordinates, movement trajectory point set and acceleration data;

[0024] According to the keyboard tapping data, a dynamic time warping algorithm execution environment is constructed. The real-time tapping sequence and each sequence in the historical tapping sequence library are compared with each other, and the warping distance is calculated according to the dynamic time warping algorithm to obtain a similarity matrix. For the mouse behavior data, based on the hidden Markov model theory, the mouse operation state space is defined, and the state transition probability matrix and observation probability matrix are estimated according to the historical data. A hidden Markov model is constructed, and the real-time mouse operation data is used as the observation sequence and input into the hidden Markov model to calculate the similarity probability of mouse operation.

[0025] The keyboard similarity matrix and the mouse operation similarity probability are used to generate a data tensor, which is input into the trained temporal convolutional network. The deviation value between the current data and the historical learning pattern is output, and the verification is judged by comparing it with the preset deviation threshold.

[0026] As a further solution of the present invention: in S5, the multi-node cross-validation includes:

[0027] Obtain copies of the abnormal data block from nodes in at least five geographical areas and compare the timestamp continuity of the verification code. If more than three copies fail to verify, it is determined that the abnormal data block has been attacked and the data tracing program is initiated.

[0028] As a further solution of the present invention: in S6, analyzing the user's behavior characteristics from accessing files to operating files, and identifying whether there is abnormal file operation includes:

[0029] The file operation behavior space is divided into several subspaces, including file preview, editing and downloading. For the operation events in each subspace, the time series is used as the dimension to collect the operation occurrence time and operation duration. For specific file formats, in the data preprocessing stage, a feature mapping matrix is ​​constructed based on the format characteristics to map the original operation data to the quantum state simulation space; for the operation data in the quantum state simulation space, the quantum correlation between different operation events is calculated. By setting multi-level threshold logic, when the quantum correlation exceeds the preset normal range threshold set, it is determined that there is a potential abnormal operation combination;

[0030] After the abnormal operation combination is determined, based on the timestamp index, the preceding and following operation sequences of each operation event in the abnormal operation combination are traced back to build a causal graph of local operation behavior. The graph traversal algorithm is used to analyze the topological structure of nodes and edges in the causal graph to determine the propagation path and impact range of the abnormal file operation.

[0031] The present invention also includes an artificial intelligence-based bidding information intelligent management system, which is used to implement the above-mentioned artificial intelligence-based bidding information intelligent management method, including:

[0032] The feature code generation module is used to collect the hardware fingerprint and network behavior pattern of the user's operating terminal during the bidding data upload stage, and generate a composite biometric feature code including keystroke rhythm, cursor movement trajectory and file access habits;

[0033] The document graph generation module is used to extract features from the uploaded bidding documents, identify document structure features, metadata timestamp sequences and version change patterns, generate document fingerprint graphs through hash algorithms, and bind and store the document fingerprint graphs with composite biometric feature codes;

[0034] The encryption storage module is used to dynamically select the parameters of the chaotic encryption algorithm to generate encrypted bidding documents according to the current network delay and packet loss rate, divide the encrypted bidding documents into data blocks of variable length, and dynamically check each data block; randomly distribute the encrypted data blocks to storage nodes with discrete geographical locations, and each storage node only stores discontinuous data fragments;

[0035] The secondary verification module is used to verify the terminal hardware fingerprint and network behavior pattern of the requester in real time during the data access phase. When a discrepancy with the composite biometric code is detected, secondary verification is triggered;

[0036] The file reconstruction module is used to reconstruct the bidding documents after the requester passes the verification. It preferentially obtains data blocks from the three nodes with the lowest latency in parallel, verifies the continuity of the time series of the checksum, initiates multi-node cross-verification for abnormal data blocks and automatically replaces them. After the reconstruction is completed, the dynamic checksums of all data blocks are updated, and the details of the replacement operation are recorded;

[0037] The abnormal operation detection module is used to analyze the user's behavior characteristics from accessing files to operating files, identify whether there are abnormal file operations, and store the detection results in the defense strategy library.

[0038] Beneficial effects of the present invention:

[0039] The present invention generates a composite biometric code by collecting the hardware fingerprint of the user's operating terminal and the network behavior pattern, thereby realizing multi-dimensional identity authentication, improving the accuracy and security of identity recognition, and preventing identity theft; dynamically selecting chaotic encryption algorithm parameters according to the real-time network status, dividing files into variable-length data blocks and attaching dynamic check codes, randomly assigning them to discrete storage nodes, encrypting the index table and regularly rotating the node position mapping to ensure data transmission and storage security; when reconstructing data access, obtaining data blocks in parallel from low-latency nodes, verifying the continuity of the check codes, cross-verifying multiple nodes for abnormal blocks, automatically replacing and recording them, and updating the check codes to ensure data integrity and accuracy; using technologies such as time convolutional networks to divide the file operation behavior space, calculating quantum correlation to identify abnormal operation combinations, constructing a causal graph to determine the abnormal propagation path and range, accurately identifying abnormal file operations, and maintaining the fairness of bidding. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The present invention will be further described below in conjunction with the accompanying drawings.

[0041] Figure 1 It is a flow chart of an intelligent bidding information management method based on artificial intelligence of the present invention;

[0042] Figure 2 It is a module schematic diagram of an artificial intelligence-based bidding information intelligent management system of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] See also Figure 1-Figure 2 As shown, the present invention is an intelligent management method for bidding information based on artificial intelligence, comprising the following steps:

[0045] S1: Generate composite biometric code: When uploading bidding data, the system will collect some unique information of the uploading device (such as computer, mobile phone), such as the hardware fingerprint of the device, and the user's online behavior pattern, such as the rhythm of keyboard typing, the trajectory of mouse movement, and the habit of opening files, etc. This information is integrated to generate a unique composite biometric code, giving the user operation a special "label".

[0046] S2: Document fingerprint generation and binding: For the uploaded bidding documents, the system will carefully check its structure to see how the contents of the documents are organized, as well as the metadata timestamp sequence of the documents, that is, the records of when the documents were created and modified, and the version change pattern. Then, through a method called hashing algorithm, these features are turned into a document fingerprint, and this fingerprint is "bound" together with the previously generated composite biometric code and stored, so that the documents can be matched with the operators.

[0047] S3: Encryption and storage: The system will dynamically select appropriate chaotic encryption algorithm parameters to encrypt the bidding documents based on the current network conditions, such as whether the network latency is high or whether there is packet loss. After encryption, the file is cut into data blocks of different lengths, and each data block is equipped with a changing checksum. Then, these encrypted data blocks are randomly placed on storage nodes in different locations, and each storage node stores non-adjacent data fragments, which can ensure the security of file storage.

[0048] S4: Verification and secondary verification: When someone wants to access data, the system will immediately check the hardware fingerprint and network behavior pattern of the device used by the person to see if it is the same as the composite biometric code stored previously. If it is different, secondary verification will be initiated to further confirm whether it is the real person.

[0049] S5: File reconstruction: After passing the verification, the system starts to reassemble the bidding documents. It will first take the data blocks from the three storage nodes with the fastest network speed and the lowest latency at the same time. After taking them, check whether the time sequence of the checksum of each data block is correct. If an abnormal data block is found, it will take copies from several nodes in different places, compare and check, and automatically replace them if there is a problem. After the file is reassembled, the checksums of all data blocks are updated, and the situation of replacing data blocks will be recorded.

[0050] S6: Abnormal behavior detection: The system will observe the entire process from the user opening the file to the file operation to see if there are any abnormal operations. For example, whether the file preview, editing, and downloading methods are different from usual. If an abnormality is found, the situation will be stored in the defense strategy library, which can be used to strengthen the security of the system in the future.

[0051] In another preferred embodiment of the present invention, in S1, the specific process of generating a composite biometric code covers multiple detailed links. The system will accurately record the time interval of the user's mouse clicks for three consecutive file upload operations, with the help of a special timestamp recording program, accurate to the millisecond level. Subsequently, these time interval data are processed using a statistical algorithm to calculate the standard deviation, which can effectively reflect the stability of the user's click rhythm. This standard deviation is then converted into a behavioral feature vector through a specific normalization formula, so that it can be compared and analyzed under a unified standard. In the document editing scenario, the system monitors the user's frequency of using the backspace key during document editing in real time, and tracks the cursor movement speed. Using the association analysis algorithm, it deeply explores the potential connection between the two, and then generates an operation habit matrix, which can fully display the user's unique operation habits when editing documents. When the user switches between multiple tasks, the system carefully analyzes the waveform pattern of memory occupancy changes through a memory monitoring tool, and extracts the peak frequency and attenuation coefficient of the memory from it. These parameters can reflect the dynamic change characteristics of the device memory when the user switches between different tasks. Finally, the behavior feature vector, the operation habit matrix, the waveform pattern and the device MAC address are combined using a hash fusion algorithm to generate a unique identification code containing an accurate timestamp. In addition, the system will immediately compare the identification code with the pre-established user historical behavior database. If the deviation exceeds the preset threshold, it will quickly trigger an abnormal behavior warning to ensure the security of the bidding data upload process.

[0052] In another preferred embodiment of the present invention, in S3, the process of dynamically selecting the parameters of the chaotic encryption algorithm is extremely rigorous. The system continuously monitors the network delay jitter, collects delay data in real time through the network monitoring module, and calculates the packet loss ratio based on this. The channel quality index can intuitively reflect the transmission quality of the current network. According to the calculated channel quality index, the system dynamically adjusts the parameter value of the Lorentz equation. As a classic equation in chaos theory, the precise adjustment of the parameters of the Lorentz equation can generate a highly random and complex chaotic sequence. Subsequently, the original file data is encrypted using the chaotic sequence, and the encryption process adopts an efficient encryption algorithm to ensure the security of the data. In order to further optimize the encrypted data transmission and storage, the encrypted data block length is set to the golden ratio of the current network throughput, so that the network bandwidth can be fully utilized and the reasonable size of the data block can be guaranteed. In addition, a chain check code is generated for each data block, and the check code not only contains the hash value of the previous data block, but also embeds a timestamp in the chain check code of each data block. In this way, effective verification of the integrity and sequentiality of the data block can be achieved.

[0053] In another preferred embodiment of the present invention, in S3, a series of security strategies are adopted to randomly distribute the encrypted data blocks to storage nodes with discrete geographical locations. In order to enhance the security of the storage nodes, each storage node is provided with a different access protocol stack, and different protocol stacks can effectively resist different types of network attacks, increasing the difficulty for attackers to obtain data. The padding bytes of random length are inserted between adjacent data blocks. This operation makes the data structure more complicated and prevents attackers from stealing data by analyzing the rules between data blocks. At the same time, the index table recording the location of the data blocks is encrypted and stored in an independent geographical isolation area to ensure that the index information is separated from the data. Even if the data storage node is attacked, it is difficult for the attacker to obtain the storage location information of the data. In addition, the system regularly rotates the mapping relationship between the actual location of the storage node and the storage node in the system, and dynamically adjusts the data distribution according to factors such as the node load and the network connection quality by real-time monitoring of the status of each storage node to ensure the efficiency and security of data storage.

[0054] In another preferred embodiment of the present invention, in S4, the secondary verification process is extremely sophisticated and scientific. First, a temporal convolutional network structure is constructed, which is carefully designed to include multiple layers of one-dimensional convolutional layers, pooling layers and fully connected layers. Among them, the one-dimensional convolutional layer can effectively extract the local features of the data in the time dimension, the pooling layer is used to reduce the data dimension, retain key information, and the fully connected layer is responsible for integrating the features of the previous layers. Subsequently, the pre-accumulated historical keyboard tapping sequence library and the historical mouse behavior data are spliced ​​strictly according to the time dimension to form a multimodal behavior data tensor. This tensor comprehensively covers the characteristic information of the user's past operation behavior in the time series, and inputs it into the temporal convolutional network. In the network training link, the historical multimodal behavior data tensor is used as the training set, and the stochastic gradient descent optimization algorithm is adopted. The algorithm calculates the gradient of the loss function by continuously and randomly selecting training data samples, and iteratively updates the network weight parameters accordingly to minimize the cross entropy loss function. After a large number of training iterations, the network can accurately learn the temporal correlation feature pattern of the multimodal behavior data, providing an accurate reference model for subsequent verification.

[0055] In the real-time data collection stage, the input device driver interface is used to synchronously collect keyboard tapping data using the underlying system driver, and the original data sequence of tapping force and time interval is accurately obtained. These data can truly reflect the user's keyboard tapping force habits and rhythm characteristics. At the same time, with the help of the mouse device API, mouse behavior data is collected, including mouse click coordinates, movement trajectory point set and acceleration data. These data fully describe the user's behavior pattern of operating the mouse.

[0056] Based on the collected data, we analyze the keyboard tapping data and build a dynamic time warping algorithm execution environment. We calculate the warping distance between the real-time tapping sequence and the sequences in the historical tapping sequence library according to the dynamic time warping algorithm. This algorithm can effectively handle the similarity measurement problem between time series of different lengths, and then obtain the similarity matrix, which intuitively shows the similarity between real-time keyboard operations and historical operations. For mouse behavior data, based on the hidden Markov model theory, we rigorously define the mouse operation state space, estimate the state transition probability matrix and observation probability matrix based on rich historical data, and build a hidden Markov model. Then, we use the real-time mouse operation data as the observation sequence and input it into the hidden Markov model to calculate the similarity probability of mouse operations, so as to quantify the similarity between the current mouse operation and the historical normal operation.

[0057] Finally, the keyboard similarity matrix and the mouse operation similarity probability are combined to generate a data tensor and input into the trained temporal convolutional network. The network outputs the deviation value between the current data and the historical learning pattern, and makes a detailed comparison with the preset deviation threshold to determine whether the verification is passed, thereby ensuring accurate verification of user identity and operation behavior.

[0058] In another preferred embodiment of the present invention, in S5, multi-node cross-validation adopts a rigorous and efficient strategy. When an abnormal data block is detected, a copy of the abnormal data block is obtained from nodes in at least five geographical areas. These nodes are distributed in different geographical locations, which can effectively avoid data verification deviations caused by regional failures or attacks. After obtaining the copy, carefully compare the timestamp continuity of the verification code. The timestamp continuity can reflect the sequential integrity of the data block during transmission and storage. If more than three copies fail to verify, based on the majority verification principle, it is determined that the abnormal data block has been attacked, and then the data tracing program is quickly started. By tracing various information such as data transmission paths and storage node access records, the source of the attack is found to ensure the security and integrity of the data.

[0059] In another preferred embodiment of the present invention, in S6, the behavior characteristics of users from accessing files to operating files are analyzed, and the process of identifying abnormal file operations is comprehensive and in-depth. First, the file operation behavior space is finely divided into several subspaces, including file preview, editing and downloading. For the operation events in each subspace, the time series is used as the dimension to accurately collect the time when the operation occurs and the duration of the operation. These time information are crucial for analyzing the consistency and rationality of user operation behavior. When facing a specific file format, in the data preprocessing stage, a feature mapping matrix is ​​constructed based on the format features. These format features may include the structural characteristics of the file, the encoding method, etc. By constructing the mapping matrix, the original operation data is accurately mapped to the quantum state simulation space. The quantum state simulation space is not a real physical space, but an abstract concept constructed based on the principles of quantum mechanics, which is used to simulate and analyze the behavior of complex systems. Based on this, the present invention deeply explores the potential relationship between user file operation behaviors.

[0060] In principle, quantum states in quantum mechanics have unique properties such as superposition and entanglement. The quantum state simulation space draws on these properties and maps the original operation data into this virtual space through specific mathematical transformations and algorithms. In this space, seemingly isolated file operation events, such as the time and duration of file preview, editing, and downloading operations, will be interrelated according to the rules of quantum mechanics.

[0061] For example, in traditional analysis methods, we may only be able to see the order of operations and simple statistical features, but in the quantum state simulation space, different operation events will be like being in a quantum entangled state, producing deep connections that are difficult to explain with classical logic. By calculating the quantum correlation of these operation data in the quantum state simulation space, we can find those abnormal operation combinations that are difficult to detect from the perspective of conventional analysis. This method greatly improves the depth and accuracy of the analysis of user operation behavior, helps to identify potential abnormal file operation behaviors more timely and accurately, and ensures the security and stability of the bidding information management system.

[0062] For the operation data in the quantum state simulation space, the quantum correlation calculation method is used to calculate the quantum correlation between different operation events. By setting multi-level threshold logic, when the quantum correlation exceeds the preset normal range threshold set, it is determined that there is a potential abnormal operation combination. This method can keenly capture operational behavior abnormalities that are difficult to detect with traditional analysis methods.

[0063] After the abnormal operation combination is determined, the preceding and following operation sequences of each operation event in the abnormal operation combination are traced along the timeline based on the timestamp index to construct a local operation behavior causal graph, which intuitively displays the causal relationship between operation events. Using the graph traversal algorithm, the topological structure of nodes and edges in the causal graph is deeply analyzed. By analyzing the operation events represented by the nodes and the causal relationship represented by the edges, the propagation path and impact range of abnormal file operations are accurately determined, providing a strong basis for taking timely measures to prevent the spread of abnormal behavior.

[0064] The present invention also includes an artificial intelligence-based bidding information intelligent management system, which is used to implement the above-mentioned artificial intelligence-based bidding information intelligent management method, including:

[0065] The feature code generation module is used to collect the hardware fingerprint and network behavior pattern of the user's operating terminal during the bidding data upload stage, and generate a composite biometric feature code including keystroke rhythm, cursor movement trajectory and file access habits;

[0066] The document graph generation module is used to extract features from the uploaded bidding documents, identify document structure features, metadata timestamp sequences and version change patterns, generate document fingerprint graphs through hash algorithms, and bind and store the document fingerprint graphs with composite biometric feature codes;

[0067] The encryption storage module is used to dynamically select the parameters of the chaotic encryption algorithm to generate encrypted bidding documents according to the current network delay and packet loss rate, divide the encrypted bidding documents into data blocks of variable length, and dynamically check each data block; randomly distribute the encrypted data blocks to storage nodes with discrete geographical locations, and each storage node only stores discontinuous data fragments;

[0068] The secondary verification module is used to verify the terminal hardware fingerprint and network behavior pattern of the requester in real time during the data access phase. When a discrepancy with the composite biometric code is detected, secondary verification is triggered;

[0069] The file reconstruction module is used to reconstruct the bidding documents after the requester passes the verification. It preferentially obtains data blocks from the three nodes with the lowest latency in parallel, verifies the continuity of the time series of the checksum, initiates multi-node cross-verification for abnormal data blocks and automatically replaces them. After the reconstruction is completed, the dynamic checksums of all data blocks are updated, and the details of the replacement operation are recorded;

[0070] The abnormal operation detection module is used to analyze the user's behavior characteristics from accessing files to operating files, identify whether there are abnormal file operations, and store the detection results in the defense strategy library.

[0071] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An intelligent management method for bidding information based on artificial intelligence, characterized in that: The following steps are involved: S1. During the bidding data upload stage, the hardware fingerprint and network behavior pattern of the user's operating terminal are collected to generate a composite biometric code including keystroke rhythm, cursor movement trajectory and file access habits; S2. Extract features from the uploaded bidding documents, identify document structure features, metadata timestamp sequence and version change pattern, generate document fingerprint map through hash algorithm, and bind and store the document fingerprint map with the composite biometric code; S3. Dynamically select chaotic encryption algorithm parameters to generate encrypted bidding documents based on current network delay and packet loss rate, and divide the encrypted bidding documents into data blocks of variable lengths, with dynamic checksums for each data block; The encrypted data blocks are randomly distributed to geographically discrete storage nodes, and each storage node only stores discontinuous data fragments; S4. During the data access phase, the requester’s terminal hardware fingerprint and network behavior pattern are verified in real time. When a mismatch with the composite biometric code is detected, secondary verification is triggered; S5. When the requester passes the verification, the bidding document is reconstructed, and data blocks are obtained in parallel from the three nodes with the lowest delay first, the continuity of the check code time series is verified, multi-node cross-verification is initiated for abnormal data blocks and automatically replaced, and the dynamic check codes of all data blocks are updated after the reconstruction is completed, and the details of the replacement operation are recorded; S6. Analyze the user's behavior characteristics from accessing files to operating files, identify whether there are abnormal file operations, and store the detection results in the defense strategy library.

2. According to the artificial intelligence-based intelligent management method for bidding information according to claim 1, it is characterized in that: In S1, the specific process of generating the composite biometric code includes: The time interval between mouse clicks of three consecutive file upload operations by the user is recorded, the standard deviation is calculated and normalized into a behavioral feature vector; the correlation characteristics between the frequency of use of the backspace key and the cursor movement speed during document editing are extracted to generate an operation habit matrix; the waveform pattern of memory usage changes during multi-task switching is analyzed to extract the peak frequency and attenuation coefficient of the memory; the behavioral feature vector, operation habit matrix, waveform pattern and device MAC address are hashed and fused to generate a unique identification code containing a timestamp; the identification code is compared with the user's historical behavior database. If the deviation exceeds the preset threshold, an abnormal behavior warning is triggered.

3. The method for intelligent management of bidding information based on artificial intelligence according to claim 1, characterized in that: In S3, the chaotic encryption algorithm parameters are dynamically selected: Monitor network delay jitter and packet loss ratio, calculate channel quality index, dynamically adjust parameter value of Lorentz equation according to channel quality index, generate chaotic sequence, encrypt original file data through chaotic sequence, set encrypted data block length to golden ratio of current network throughput; generate a chain check code for each data block, the chain check code includes hash value of previous data block, and embed a timestamp into the chain check code of each data block.

4. The method for intelligent management of bidding information based on artificial intelligence according to claim 1, characterized in that: In S3, randomly allocating the encrypted data blocks to geographically discrete storage nodes includes: Each storage node sets a different access protocol stack, inserts padding bytes of random length between adjacent data blocks, encrypts the index table that records the location of data blocks, and stores it in an independent geographical isolation area to ensure that the index information is separated from the data; the actual location of the storage node and the mapping relationship of the storage node in the system are rotated regularly, the status of each storage node is monitored in real time, and the data distribution is adjusted dynamically.

5. The method for intelligent management of bidding information based on artificial intelligence according to claim 1, characterized in that: In S4, the secondary verification includes: Construct a temporal convolutional network structure, which includes multiple one-dimensional convolutional layers, pooling layers, and fully connected layers. The historical keyboard tapping sequence library and the historical mouse behavior data are spliced ​​into a multimodal behavior data tensor according to the time dimension and input into the temporal convolutional network. In the network training phase, the historical multimodal behavior data tensor is used as the training set, and the stochastic gradient descent optimization algorithm is used to minimize the cross entropy loss function, iteratively update the network weight parameters, and train to obtain the temporal correlation feature pattern of the multimodal behavior data; Through the input device driver interface, keyboard tapping data is synchronously collected to obtain the original data sequence of tapping force and time interval; using the mouse device API, mouse behavior data is collected, including mouse click coordinates, movement trajectory point set and acceleration data; According to the keyboard tapping data, a dynamic time warping algorithm execution environment is constructed. The real-time tapping sequence and each sequence in the historical tapping sequence library are compared with each other, and the warping distance is calculated according to the dynamic time warping algorithm to obtain a similarity matrix. For the mouse behavior data, based on the hidden Markov model theory, the mouse operation state space is defined, and the state transition probability matrix and observation probability matrix are estimated according to the historical data. A hidden Markov model is constructed, and the real-time mouse operation data is used as the observation sequence and input into the hidden Markov model to calculate the similarity probability of mouse operation. The keyboard similarity matrix and the mouse operation similarity probability are used to generate a data tensor, which is input into the trained temporal convolutional network. The deviation value between the current data and the historical learning pattern is output, and the verification is judged by comparing it with the preset deviation threshold.

6. The method for intelligent management of bidding information based on artificial intelligence according to claim 1, characterized in that: In S5, the multi-node cross-validation includes: Obtain copies of the abnormal data block from nodes in at least five geographical areas and compare the timestamp continuity of the verification code. If more than three copies fail to verify, it is determined that the abnormal data block has been attacked and the data tracing program is initiated.

7. The method for intelligent management of bidding information based on artificial intelligence according to claim 1, characterized in that: In S6, analyzing the user's behavior characteristics from accessing the file to operating the file, and identifying whether there is an abnormal file operation includes: The file operation behavior space is divided into several subspaces, including file preview, editing and downloading. For the operation events in each subspace, the time series is used as the dimension to collect the operation occurrence time and operation duration. For specific file formats, in the data preprocessing stage, a feature mapping matrix is ​​constructed based on the format characteristics to map the original operation data to the quantum state simulation space; for the operation data in the quantum state simulation space, the quantum correlation between different operation events is calculated. By setting multi-level threshold logic, when the quantum correlation exceeds the preset normal range threshold set, it is determined that there is a potential abnormal operation combination; After the abnormal operation combination is determined, based on the timestamp index, the preceding and following operation sequences of each operation event in the abnormal operation combination are traced back to build a causal graph of local operation behavior. The graph traversal algorithm is used to analyze the topological structure of nodes and edges in the causal graph to determine the propagation path and impact range of the abnormal file operation.

8. An intelligent bidding information management system based on artificial intelligence, used to implement an intelligent bidding information management method based on artificial intelligence according to any one of claims 1 to 7, characterized in that: include: The feature code generation module is used to collect the hardware fingerprint and network behavior pattern of the user's operating terminal during the bidding data upload stage, and generate a composite biometric feature code including keystroke rhythm, cursor movement trajectory and file access habits; The document graph generation module is used to extract features from the uploaded bidding documents, identify document structure features, metadata timestamp sequences and version change patterns, generate document fingerprint graphs through hash algorithms, and bind and store the document fingerprint graphs with composite biometric feature codes; The encryption storage module is used to dynamically select the parameters of the chaotic encryption algorithm to generate an encrypted bidding document according to the current network delay and packet loss rate, and to divide the encrypted bidding document into data blocks of variable lengths, and each data block has a dynamic checksum; The encrypted data blocks are randomly distributed to geographically discrete storage nodes, and each storage node only stores discontinuous data fragments; The secondary verification module is used to verify the terminal hardware fingerprint and network behavior pattern of the requester in real time during the data access phase. When a discrepancy with the composite biometric code is detected, secondary verification is triggered; The file reconstruction module is used to reconstruct the bidding documents after the requester passes the verification. It preferentially obtains data blocks from the three nodes with the lowest latency in parallel, verifies the continuity of the time series of the checksum, initiates multi-node cross-verification for abnormal data blocks and automatically replaces them. After the reconstruction is completed, the dynamic checksums of all data blocks are updated, and the details of the replacement operation are recorded; The abnormal operation detection module is used to analyze the user's behavior characteristics from accessing files to operating files, identify whether there are abnormal file operations, and store the detection results in the defense strategy library.

Images