A security-enhanced data verification method and system for intelligent platforms

Abstract

This invention provides a security-enhanced data verification method and system for intelligent platforms. The method includes: storing network attack data on a blockchain and connecting each data block node using a circular linked list; optimizing data imbalance during data preprocessing using a multi-branch path tree balancing method; training a model by integrating multi-source heterogeneous attack data, and integrating data sources using a circular linked list and a multi-branch path tree; recording parameter update processes on the blockchain and managing different versions of parameters using a circular linked list during AI model training; recording inference requests and results on the blockchain during the inference phase, tracking inference task relationships using a circular linked list and a multi-branch path tree, and verifying the correctness of inference results; evaluating model performance based on blockchain-recorded inference results, and recording abnormal branch access paths during the inference process for adversarial attacks. This invention combines the advantages of blockchain, circular linked lists, and multi-branch path trees to optimize AI model data verification and processing, improve the model's performance in identifying network attacks, and is of great significance for network security protection.

Description

Technical Field

[0001] This application relates to the field of network security technology, and in particular to a security-enhanced data verification method and system for intelligent platforms. Background Technology

[0002] With the rapid development of network technology, network attack methods are becoming increasingly complex and diverse, posing a serious threat to network security. AI models have great potential in identifying network attacks, but their performance and reliability largely depend on the integrity of the data used in the training and inference processes. Traditional data processing and verification methods have many shortcomings when facing the dynamic, complex, and unbalanced nature of network attack data, such as low efficiency in data integrity verification and poor adaptability of models to new types of attacks.

[0003] Therefore, there is an urgent need for a targeted, security-enhanced data verification method and system for intelligent platforms. Summary of the Invention

[0004] The purpose of this invention is to provide a security-enhanced data verification method and system for intelligent platforms, which improves the efficiency of data integrity verification and model performance of AI models when identifying network attacks by optimizing data structure and verification process.

[0005] In a first aspect, this application provides a security-enhanced data verification method for intelligent platforms, the method comprising:

[0006] Network attack data is stored on the blockchain, with each data block's storage location on the blockchain serving as a node, and these nodes are connected through a circular linked list;

[0007] When performing data cleaning and normalization preprocessing operations, the balancing method of multi-branch path tree is used to treat the data imbalance that occurs during the data preprocessing process as a path tree structure under imbalance, and optimize and balance it to make the data of each feature dimension tend to be balanced.

[0008] The training model integrates multi-source heterogeneous attack data, utilizes circular linked lists to efficiently integrate the data storage locations of each source, and quickly reads samples from different data sources as needed during training. It also balances the proportion of data from each source through a multi-branch path tree.

[0009] During AI model training, the update process of model parameters is recorded on the blockchain, and a circular linked list is used to manage the relationship between different versions of model parameters.

[0010] Each time the parameters are updated, the new parameters are added as a node to the circular linked list and verified through the blockchain's consensus mechanism.

[0011] During the AI ​​model inference stage, inference requests and results are recorded on the blockchain. A circular linked list is used to track the relationship between multiple inference requests and results of the same task in a time sequence. At the same time, a multi-branch path tree is used to record the association and dependency relationships between different inference tasks.

[0012] When it is necessary to verify the correctness of the reasoning result, the data and integrity verification scheme on the blockchain are used to compare the reasoning process and result of this reasoning request with the reasoning process and result of similar previous requests, and combined with the model version and data source information recorded in the path tree, to determine whether the current reasoning result is reasonable.

[0013] Based on blockchain-recorded reasoning results, a multi-branch path tree is constructed to evaluate model performance.

[0014] To counter adversarial attacks, a multi-branch path tree is used to record abnormal branch access paths during the inference process.

[0015] Secondly, this application provides a security-enhanced data verification system for intelligent platforms, the system comprising:

[0016] The data storage module is used to store network attack data on the blockchain. Each data block is stored at a node on the blockchain, and these nodes are connected by a circular linked list.

[0017] The preprocessing module is used to perform data cleaning and normalization preprocessing operations. It uses the balancing method of multi-branch path tree to treat the data imbalance that occurs during the data preprocessing process as a path tree structure under imbalance, and optimizes and balances it to make the data of each feature dimension tend to be balanced.

[0018] The training module is used to train models by fusing multi-source heterogeneous attack data. It uses a circular linked list to efficiently integrate the data storage locations of each source, and reads samples from different data sources quickly as needed during training. It balances the proportion of each source data through a multi-branch path tree.

[0019] The verification module is used to record the update process of model parameters on the blockchain during AI model training, and uses a circular linked list to manage the relationship between different versions of model parameters. After each parameter update, the new parameter is added as a node to the circular linked list and verified through the blockchain's consensus mechanism.

[0020] The inference module is used to record inference requests and results on the blockchain during the AI ​​model inference stage. It uses a circular linked list to track the relationship between multiple inference requests and results of the same task in a time sequence, and records the association and dependency relationships between different inference tasks through a multi-branch path tree.

[0021] When it is necessary to verify the correctness of the reasoning result, the data and integrity verification scheme on the blockchain are used to compare the reasoning process and result of this reasoning request with the reasoning process and result of similar previous requests, and combined with the model version and data source information recorded in the path tree, to determine whether the current reasoning result is reasonable.

[0022] The evaluation module is used to construct a multi-branch path tree to evaluate the performance of the model based on the inference results recorded on the blockchain.

[0023] The recording module is used to record abnormal branch access paths during the inference process using a multi-branch path tree to counter adversarial attacks.

[0024] Thirdly, this application provides a security-enhanced data verification system for intelligent platforms, the system comprising a processor and a memory:

[0025] The memory is used to store program code and transmit the program code to the processor;

[0026] The processor is configured to execute any one of the methods described in the first aspect according to the instructions in the program code.

[0027] Fourthly, this application provides a computer-readable storage medium for storing program code, which is executed by a processor to implement the method described in any one of the first aspects.

[0028] Beneficial effects

[0029] This invention provides a security-enhanced data verification method and system for intelligent platforms. The method includes: storing network attack data on a blockchain and connecting each data block node with a circular linked list; optimizing data imbalance during data preprocessing using a multi-branch path tree balancing method; training a model by integrating multi-source heterogeneous attack data and integrating data sources using a circular linked list and a multi-branch path tree; recording parameter update processes on the blockchain and managing different versions of parameters using a circular linked list during AI model training; recording inference requests and results on the blockchain during the inference phase, tracking inference task relationships using a circular linked list and a multi-branch path tree, and verifying the correctness of inference results; evaluating model performance based on blockchain-recorded inference results, and recording abnormal branch access paths during the inference process for adversarial attacks.

[0030] The method and system of the present invention have the following advantages and effects:

[0031] By applying AI data verification and processing methods based on blockchain, circular linked lists, and multi-branch path trees to identify network attacks, the efficiency of data integrity verification in the process of identifying network attacks by AI models is effectively improved, ensuring the accuracy and reliability of the data used for model training and inference.

[0032] By using a multi-branch path tree balancing method to optimize the data preprocessing and model inference process, the AI ​​model can better cope with the imbalance and complexity of network attack data, and improve the model's ability to identify various network attacks, especially new hybrid attacks and adversarial attacks. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0034] Figure 1 This is a flowchart of the present invention;

[0035] Figure 2 This is a system architecture diagram of the present invention. Detailed Implementation

[0036] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention.

[0037] AI models hold immense potential in identifying cyberattacks, but their performance and reliability largely depend on the integrity of the data used during training and inference. Traditional data processing and verification methods suffer from numerous shortcomings when dealing with the dynamic, complex, and unbalanced nature of cyberattack data, such as low efficiency in data integrity verification and poor model adaptability to novel attacks.

[0038] By optimizing data structures and verification processes, the efficiency of AI models in verifying data integrity and the performance of models when identifying network attacks can be improved.

[0039] This application provides a security-enhanced data verification method for intelligent platforms, the method comprising:

[0040] Network attack data is stored on the blockchain, with each data block's storage location on the blockchain serving as a node, and these nodes are connected through a circular linked list;

[0041] When performing data cleaning and normalization preprocessing operations, the balancing method of multi-branch path tree is used to treat the data imbalance that occurs during the data preprocessing process as a path tree structure under imbalance, and optimize and balance it to make the data of each feature dimension tend to be balanced.

[0042] The training model integrates multi-source heterogeneous attack data, utilizes circular linked lists to efficiently integrate the data storage locations of each source, and quickly reads samples from different data sources as needed during training. It also balances the proportion of data from each source through a multi-branch path tree.

[0043] During AI model training, the update process of model parameters is recorded on the blockchain, and a circular linked list is used to manage the relationship between different versions of model parameters.

[0044] Each time the parameters are updated, the new parameters are added as a node to the circular linked list and verified through the blockchain's consensus mechanism.

[0045] During the AI ​​model inference stage, inference requests and results are recorded on the blockchain. A circular linked list is used to track the relationship between multiple inference requests and results of the same task in a time sequence. At the same time, a multi-branch path tree is used to record the association and dependency relationships between different inference tasks.

[0046] When it is necessary to verify the correctness of the reasoning result, the data and integrity verification scheme on the blockchain are used to compare the reasoning process and result of this reasoning request with the reasoning process and result of similar previous requests, and combined with the model version and data source information recorded in the path tree, to determine whether the current reasoning result is reasonable.

[0047] Based on blockchain-recorded reasoning results, a multi-branch path tree is constructed to evaluate model performance.

[0048] To counter adversarial attacks, a multi-branch path tree is used to record abnormal branch access paths during the inference process.

[0049] Analyze the perturbation patterns of input sample features corresponding to these abnormal paths, adjust the weight thresholds of key features in the model, and enhance the robustness of the model.

[0050] The effectiveness of identifying different attack types is evaluated from multiple dimensions, including accuracy, recall, and F1 score, and performance metrics are used as weights for nodes in a multi-branch path tree. For branches with poor performance, the corresponding training data and model parameters are analyzed backtrackingly, and targeted optimization strategies are adjusted, such as supplementing relevant attack data and fine-tuning model parameters, forming a closed loop for continuous model optimization.

[0051] In some preferred embodiments, during training, the differences between new and old parameters are compared, and the learning rate and optimization direction are reasonably adjusted based on historical update results to enhance the model's ability to identify novel hybrid attacks.

[0052] In some preferred embodiments, a multi-branch path tree balancing method is used to construct a multi-branch path tree for the inference task according to the attack type and urgency, dynamically monitor the load of each branch, reasonably allocate more computing resources to high-load branches, optimize the scheduling strategy, and enable the model to efficiently process various attack detection tasks with limited resources.

[0053] In complex scenarios, a combination of circular linked lists and multi-branch path trees is used for integrity verification. Efficient hash value verification is performed along the branches of the circular linked lists and path trees to ensure that the AI ​​model uses complete and accurate data.

[0054] In some preferred embodiments, the data is constructed into a multi-branch path tree according to the feature dimension, the amount of data in each branch is dynamically monitored, and unbalanced branches are adjusted by oversampling, undersampling or feature transformation to make the data in each feature dimension tend to be balanced, thereby improving the model's ability to perceive various attack features.

[0055] The parameter update path during model training is optimized using a multi-branch path tree balancing method. If an imbalance in parameter updates occurs during training, it is considered an imbalanced state of the path tree. The optimized balancing method is used to adjust the parameter update rhythm, making the model converge more stably during training.

[0056] The system may further include the following process: the verifier issues a challenge request, performs integrity verification, and feeds back the verification result to the data owner.

[0057] Cloud storage servers store data and generate auxiliary information to return to the blockchain.

[0058] Blockchain needs to store data integrity verification information, forward integrity verification challenge requests, and return auxiliary information returned by the cloud storage server as evidence to the verifier.

[0059] Figure 2 The architecture diagram of the security-enhanced data verification system applied to an intelligent platform provided in this application includes:

[0060] The data storage module is used to store network attack data on the blockchain. Each data block is stored at a node on the blockchain, and these nodes are connected by a circular linked list.

[0061] The preprocessing module is used to perform data cleaning and normalization preprocessing operations. It uses the balancing method of multi-branch path tree to treat the data imbalance that occurs during the data preprocessing process as a path tree structure under imbalance, and optimizes and balances it to make the data of each feature dimension tend to be balanced.

[0062] The training module is used to train models by fusing multi-source heterogeneous attack data. It uses a circular linked list to efficiently integrate the data storage locations of each source, and reads samples from different data sources quickly as needed during training. It balances the proportion of each source data through a multi-branch path tree.

[0063] The verification module is used to record the update process of model parameters on the blockchain during AI model training, and uses a circular linked list to manage the relationship between different versions of model parameters. After each parameter update, the new parameter is added as a node to the circular linked list and verified through the blockchain's consensus mechanism.

[0064] The inference module is used to record inference requests and results on the blockchain during the AI ​​model inference stage. It uses a circular linked list to track the relationship between multiple inference requests and results of the same task in a time sequence, and records the association and dependency relationships between different inference tasks through a multi-branch path tree.

[0065] When it is necessary to verify the correctness of the reasoning result, the data and integrity verification scheme on the blockchain are used to compare the reasoning process and result of this reasoning request with the reasoning process and result of similar previous requests, and combined with the model version and data source information recorded in the path tree, to determine whether the current reasoning result is reasonable.

[0066] The evaluation module is used to construct a multi-branch path tree to evaluate the performance of the model based on the inference results recorded on the blockchain.

[0067] The recording module is used to record abnormal branch access paths during the inference process using a multi-branch path tree to counter adversarial attacks.

[0068] This application provides a security-enhanced data verification system for intelligent platforms, the system comprising: a processor and a memory.

[0069] The memory is used to store program code and transmit the program code to the processor;

[0070] The processor is configured to execute the method described in any one of the embodiments of the first aspect according to the instructions in the program code.

[0071] This application provides a computer-readable storage medium for storing program code, which is executed by a processor to implement the method described in any one of the embodiments of the first aspect.

[0072] In a specific implementation, the present invention also provides a computer storage medium, wherein the computer storage medium may store a program, and the program, when executed, may include some or all of the steps in the various embodiments of the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0073] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0074] The same or similar parts between the various embodiments in this specification can be referred to mutually. In particular, the embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the description in the method embodiments.

[0075] The embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention.

Claims

1. A security enhanced data validation method applied to a smart platform, characterized in that, The method comprises: Storing network attack data on a blockchain, each data block being stored on the blockchain as a node, and connecting the nodes through a circular linked list; When performing cleaning and normalization preprocessing operations on the data, a balanced method of a multi-branch path tree is used, data imbalance occurring during data preprocessing is regarded as a path tree structure in an imbalanced state, and the path tree structure is optimized and balanced, a multi-branch path tree is constructed according to feature dimensions, the amount of data in each branch is dynamically monitored, and for an imbalanced branch, oversampling, undersampling, or feature transformation adjustment is performed to make the data in each feature dimension tend to be balanced; A model is trained by fusing multi-source heterogeneous attack data, different data source samples are quickly read on demand during training by using the circular linked list to integrate the storage locations of the source data, and the proportion of each source data is balanced by using the multi-branch path tree; During AI model training, the updating process of model parameters is recorded on the blockchain, and the relationship between different versions of model parameters is managed by using the circular linked list; After each parameter update, the new parameters are added as a node to the circular linked list, and the new parameters are verified by using the consensus mechanism of the blockchain; During the inference stage of the AI model, inference requests and results are recorded on the blockchain, the relationship between multiple inference requests and results of the same task is tracked in time sequence by using the circular linked list, and the association and dependency relationship between different inference tasks are recorded by using the multi-branch path tree; When the correctness of the inference result needs to be verified, the data and integrity verification scheme on the blockchain are used to compare the inference process and result of the current inference request with those of a similar request in the past, and the model version and data source information recorded in the path tree are used to determine whether the current inference result is reasonable; Based on the recorded inference result on the blockchain, a multi-branch path tree evaluation model is constructed, and the performance indicators of the model in each dimension are used as the node weights of the multi-branch path tree; For adversarial attacks, the multi-branch path tree is used to record abnormal branch access paths during the inference process.

2. The method of claim 1, wherein: During training, the difference between new and old parameters is compared, the learning rate and optimization direction are adjusted reasonably in combination with the historical updating effect, and the recognition performance of the model for new hybrid attacks is enhanced.

3. The method of claim 2, wherein: By using the multi-branch path tree balancing method, inference tasks are constructed into a multi-branch path tree according to attack types and urgency, the load of each branch is dynamically monitored, more computing resources are allocated to high-load branches, the scheduling strategy is optimized, and the model can process various attack detection tasks under limited resources.

4. A security enhanced data validation system for smart platforms, characterized in that, The system comprises: A data storage module is configured to store network attack data on a blockchain, each data block being stored on the blockchain as a node, and connecting the nodes through a circular linked list; A preprocessing module is configured to, when performing cleaning and normalization preprocessing operations on the data, use a balanced method of a multi-branch path tree, regard data imbalance occurring during data preprocessing as a path tree structure in an imbalanced state, and optimize and balance the path tree structure, construct a multi-branch path tree according to feature dimensions, dynamically monitor the amount of data in each branch, and for an imbalanced branch, perform oversampling, undersampling, or feature transformation adjustment to make the data in each feature dimension tend to be balanced. The training module is used for training a model by fusing multi-source heterogeneous attack data, integrating storage positions of various source data by using a circular linked list, quickly reading different data source samples on demand during training, and balancing proportions of various source data by using a multi-branch path tree; The verification module is used for recording an updating process of model parameters on a blockchain during AI model training, managing relationships between model parameters of different versions by using a circular linked list, adding new parameters as a node to the circular linked table after each parameter update, and verifying the new parameters by using a consensus mechanism of the blockchain; The reasoning module is used for recording a reasoning request and a result on the blockchain during an AI model reasoning stage, tracking relationships between multiple reasoning requests and results of the same task in a time sequence by using a circular linked list, and recording association and dependency relationships between different reasoning tasks by using a multi-branch path tree; When it is necessary to verify correctness of a reasoning result, data on the blockchain and an integrity verification scheme are used to compare a current reasoning request with a reasoning process and a result of a similar request in the past, and to judge whether the current reasoning result is reasonable in combination with model version and data source information recorded in the path tree; The evaluation module is used for recording a reasoning result based on the blockchain, constructing a multi-branch path tree evaluation model, and taking performance indicators of the model in various dimensions as node weights of the multi-branch path tree; The recording module is used for recording an abnormal branch access path in a reasoning process by using a multi-branch path tree for an adversarial attack.

5. A security enhanced data verification system for smart platforms, characterized in that, The system includes a processor and a memory: The memory is used for storing program code and transmitting the program code to the processor; The processor is used for executing instructions in the program code to implement the method in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium is used for storing program code, and the program code is used for being executed by the processor to implement the method in any one of claims 1-3.

Images