Block chain encryption and deep learning retrieval method for meta universe e-commerce platform

Abstract

The invention discloses a block chain encryption and deep learning retrieval method for a meta universe e-commerce platform, and the method comprises the steps: carrying out the privacy protection of user behavior data, virtual commodity features and social network relationships through a homomorphic encryption technology and a zero-knowledge proof protocol, generating a transaction voucher which cannot be tampered, and constructing a multi-mode encryption data set; designing a deep learning retrieval model with an attention mechanism and optimizing parameters of the deep learning retrieval model; a lightweight model is generated through quantitative pruning processing, the lightweight model is deployed at distributed edge nodes, and real-time rendering and low-delay interaction of virtual commodities are realized in combination with a dynamic resource scheduling algorithm; introducing a federated learning framework to carry out incremental updating on the model; a cloud edge collaborative block chain synchronization mechanism is established, and data consistency is ensured by using a fragmentation technology and a consensus algorithm. According to the method, the problems of privacy leakage, insufficient 3D feature analysis, high high-concurrency interaction delay, low data collaboration efficiency and the like of a traditional meta universe e-commerce platform are solved, and the comprehensive performance of the platform is remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the cross - technical field of e - commerce and information technology, and specifically relates to a method for blockchain encryption and deep - learning retrieval for a metaverse e - commerce platform. Background Art

[0002] With the rapid development of metaverse technology, metaverse e - commerce platforms have gradually become an important carrier integrating virtual and real economies. Users conduct product browsing, real - time transactions, and social activities through immersive interactions, which pose multi - dimensional challenges to the platform's technical architecture. However, existing technical solutions have significant defects in core areas such as privacy protection, feature analysis, resource scheduling, and data collaboration, severely restricting the availability, security, and scalability of the platform. Traditional metaverse e - commerce platforms usually adopt a centralized data storage and plain - text processing method, making user behavior data, social relationships, and transaction information extremely vulnerable to being stolen or tampered with by malicious attackers. Although some platforms attempt to introduce blockchain technology to ensure the immutability of transaction records, they lack encryption protection for the original data. For example, sensitive information such as user browsing preferences and 3D product features may still be reverse - cracked through data correlation analysis. In addition, frequent data decryption operations during transaction verification further increase the risk of privacy leakage. Existing deep - learning models have difficulty effectively capturing key features when processing multi - modal data (such as 3D models, textures, materials). Traditional retrieval models mostly perform matching based on Euclidean distance or cosine similarity, unable to distinguish local detail differences in complex 3D structures, resulting in low accuracy of product recommendations. At the same time, optimization strategies for multi - modal data, such as attention mechanisms, are not introduced during the model training process, further restricting the efficiency of feature analysis. Metaverse e - commerce platforms need to support a large number of users to simultaneously perform virtual product rendering and real - time interactions. Existing solutions mostly rely on cloud - centralized computing, resulting in long data transmission paths and significant rendering delays. Although edge computing technology is partially adopted, it does not combine lightweight model deployment with dynamic resource scheduling strategies, and the problem of uneven computing loads on edge nodes is prominent, making it difficult to meet the low - latency interaction requirements. The platform needs to synchronize user data, model parameters, and transaction records between the cloud and edge nodes. Traditional blockchain technology has low synchronization efficiency due to large amounts of data on the chain and complex consensus mechanisms. In addition, model updates mostly adopt a centralized training mode, requiring uploading original data to the cloud, which poses a risk of privacy leakage, and there is a lack of efficient framework support for cross - node collaborative training. Summary of the Invention

[0003] The present invention aims to provide a method for blockchain encryption and deep - learning retrieval for a metaverse e - commerce platform, which can solve the above problems.

[0004] To solve the above problems, the technical solutions adopted by the present invention are as follows: A method for blockchain encryption and deep learning retrieval in a metaverse e-commerce platform, comprising: using homomorphic encryption technology to protect the privacy of user behavior data, virtual commodity features, and social network relationships, and combining with the zero-knowledge proof protocol to generate tamper-proof transaction vouchers, and constructing an encrypted dataset containing multi-modal features; designing a deep learning retrieval model, enhancing the parsing ability of 3D commodity features by introducing an attention mechanism, and optimizing the model parameters using a contrast loss function; performing quantization pruning on the deep learning retrieval model to obtain a lightweight retrieval model, and deploying the lightweight retrieval model to distributed edge nodes, and combining with a dynamic resource scheduling algorithm to achieve real-time rendering and low-latency interaction of virtual commodities; on the edge nodes, introducing a federated learning framework to incrementally update the lightweight retrieval model, and through the way of local model training and global model aggregation, continuously optimize the model on the premise of ensuring data privacy; establishing a cloud-edge collaborative blockchain synchronization mechanism, using sharding technology to reduce the amount of data transmission on the chain, and ensuring data consistency between the cloud and edge nodes through a consensus algorithm.

[0005] Specifically, the using of homomorphic encryption technology to protect the privacy of user behavior data, virtual commodity features, and social network relationships includes: Obtaining the original data of user behavior data, virtual commodity features, and social network relationships; Inputting the original data into a homomorphic encryption module to generate an encrypted dataset; Combining with the zero-knowledge proof protocol to generate tamper-proof transaction vouchers, and storing the transaction vouchers in the blockchain.

[0006] Specifically, the process of designing the deep learning retrieval model includes: Constructing a basic deep learning model, the basic deep learning model including an input layer, a hidden layer, and an output layer; Introducing an attention mechanism module in the hidden layer for enhancing the parsing ability of 3D commodity features; Defining a contrast loss function for optimizing the model parameters.

[0007] Specifically, the obtaining of a lightweight retrieval model by performing quantization pruning on the deep learning retrieval model includes: Obtaining the trained deep learning retrieval model; Adopting model pruning technology to remove redundant parameters, and using a dynamic range quantization method to quantize floating-point weights into fixed-bit-width integers; Performing performance testing on the quantized model, and adjusting the pruning and quantization parameters according to the test results to ensure the stability of the model performance.

[0008] Specifically, the dynamic resource scheduling algorithm includes: Obtain the real-time load information of edge nodes, including computing requirements, network latency, and node utilization; Calculate the resource allocation priority according to the real-time load information; Dynamically adjust the task allocation strategy of edge nodes according to the resource allocation priority.

[0009] Specifically, the incremental update of the lightweight retrieval model by introducing the federated learning framework includes: On edge nodes, locally train the lightweight retrieval model using local data; Upload the locally trained model parameters to the cloud; In the cloud, aggregate the locally trained model parameters uploaded by multiple edge nodes to generate a global model; Send the global model to each edge node to complete model update.

[0010] Specifically, the establishment of a cloud-edge collaborative blockchain synchronization mechanism includes: Divide the on-chain data into multiple shards, each shard containing part of the data; Calculate the data transmission time after sharding; Use the consensus algorithm to verify the sharded data to ensure data consistency between the cloud and edge nodes.

[0011] Specifically, the real-time rendering and low-latency interaction of virtual goods include: Obtain the virtual good data requested by the user; Input the virtual good data into the lightweight retrieval model to generate the feature vectors required for rendering; According to the feature vectors, call the rendering engine of the edge node for real-time rendering; Return the rendering result to the user terminal to complete low-latency interaction.

[0012] Specifically, the zero-knowledge proof protocol generates an immutable transaction voucher, including: Obtain the transaction data and perform a hash operation on it to generate a transaction digest; Use the zero-knowledge proof algorithm to verify the transaction digest to generate a transaction voucher; Store the transaction voucher in the blockchain to ensure the immutability of the transaction data.

[0013] Specifically, the construction of an encrypted dataset containing multi-modal features includes: Obtain the original data of user behavior data, virtual good features, and social network relationships; Preprocess the original data, including data cleaning, feature extraction, and normalization operations; Input the preprocessed data into the homomorphic encryption module to generate an encrypted multi-modal feature dataset.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: Through homomorphic encryption and zero-knowledge proof protocol, the encrypted storage of user data and the non-tampering of transaction vouchers are realized, effectively preventing privacy leakage and transaction fraud; the deep learning model introducing the attention mechanism is optimized in combination with the contrast loss function, significantly improving the parsing accuracy and retrieval efficiency of multi-modal 3D commodity features; the lightweight model is deployed on the edge node, combined with the dynamic resource allocation strategy, reducing the computing load and realizing the real-time rendering and low-latency interaction of virtual commodities; the federated learning framework supports local training and global aggregation of edge nodes, realizing model iterative update without sharing the original data; the cloud-edge collaborative blockchain sharding technology reduces the amount of data transmitted on the chain, and combines the consensus algorithm to ensure the efficiency and consistency of cross-node data synchronization.

[0015] To make the above objects, features, and advantages of the present invention more obvious and understandable, specific embodiments of the present invention are hereinafter given, and in conjunction with the accompanying drawings, the detailed description is as follows. Description of the Drawings

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a flowchart of the method of the present invention, including the encryption of user behavior data, the generation of transaction vouchers by the zero-knowledge proof protocol, and the construction process of the encrypted multi-modal feature dataset.

[0018] Figure 2 It is a schematic diagram of the cloud-edge collaborative blockchain synchronization mechanism, showing the application of sharding technology in on-chain data transmission, the incremental update process of the federated learning framework, and the operation logic of the dynamic resource scheduling algorithm. Detailed Embodiments

[0019] The present invention provides a method for blockchain encryption and deep learning retrieval for a metaverse e-commerce platform. The following combines Figure 1 and Figure 2Describe its specific implementation details. This embodiment takes the actual application scenario of a large-scale metaverse e-commerce as an example to show how to achieve user privacy protection, accurate virtual commodity recommendation, and high-concurrency and low-latency interaction through homomorphic encryption technology, zero-knowledge proof protocol, improved deep learning retrieval model, quantization pruning processing, dynamic resource scheduling algorithm, federated learning framework, and cloud-edge collaborative blockchain synchronization mechanism.

[0020] Construct an encrypted dataset containing multi-modal features: Obtain the original data of user behavior data, virtual commodity features, and social network relationships. In actual applications, this data comes from the operation records of users in the metaverse platform, including behaviors such as clicks, browsing, and purchases, as well as feature data such as 3D model files, material parameters, and texture information of virtual commodities, and also includes social interaction records between users.

[0021] Preprocess the original data, including data cleaning, feature extraction, and normalization operations. For example, for user behavior data, eliminate invalid or duplicate operation records, and convert discrete features into numerical forms through one-hot encoding; for virtual commodity features, extract the key point coordinates, normal vector distribution, and texture mapping matrix of the 3D model; for social network relationships, use an adjacency matrix to represent the connection strength between users.

[0022] Input the preprocessed data into the homomorphic encryption module to generate an encrypted multi-modal feature dataset. The homomorphic encryption module uses the Paillier encryption algorithm, which allows direct addition and multiplication operations on encrypted data without revealing the original data, thus avoiding the risk of privacy leakage that may occur during the decryption process. Further, combine the zero-knowledge proof protocol to generate an immutable transaction voucher.

[0023] Obtain transaction data and perform a hash operation on it to generate a transaction digest.

[0024] Use the zero-knowledge proof algorithm to verify the transaction digest and generate a transaction voucher. Specifically, the transaction data includes commodity order information, payment amount, and timestamp. Generate a fixed-length transaction digest through the SHA-256 algorithm, and then use the zero-knowledge proof algorithm to verify the authenticity of the transaction and generate a transaction voucher that only contains proof information without exposing the specific content.

[0025] Store the transaction voucher in the blockchain to ensure the immutability of transaction data. As Figure 1 shown, this module realizes the complete process from original data to encrypted data and then to immutable transaction vouchers, providing a secure and reliable data foundation for subsequent modules.

[0026] Design and optimization of the deep learning retrieval model: Build a basic deep learning model, which includes an input layer, a hidden layer, and an output layer. The input layer receives the encrypted multi-modal feature dataset. The hidden layer consists of multiple fully connected neural networks, and the output layer is responsible for generating the final feature vector.

[0027] Introduce an attention mechanism module in the hidden layer to enhance the ability to capture the key features of 3D products. Specifically, the attention mechanism module calculates the importance weights of each feature point and preferentially focuses on the key areas of 3D products, such as shape contours, material reflectivity, and texture details.

[0028] Define a contrastive loss function to optimize the model parameters. The formula is: Where, represents the feature vector extracted by the deep learning model, is the Euclidean distance function, represents the label indicating whether the sample pair is similar. m is a preset boundary value used to distinguish the distances of positive and negative sample pairs. During the training process, by minimizing the contrastive loss function, the feature vectors of similar samples are made closer, and the feature vectors of dissimilar samples are made farther apart, thus significantly improving the retrieval accuracy and generalization ability.

[0029] Quantization and pruning processing of the lightweight retrieval model: Obtain the trained deep learning retrieval model.

[0030] Adopt model pruning technology to remove redundant parameters and use dynamic range quantization method to quantize floating-point weights into fixed-bit integer. Model pruning technology analyzes the importance of weights and removes parameters that contribute less to the model performance, such as weight values below a certain threshold. The dynamic range quantization method determines the quantization interval and bit width through statistical analysis of the weight distribution, such as quantizing 32-bit floating-point numbers into 8-bit integers, thereby reducing the computational complexity and storage requirements of the model.

[0031] Conduct performance testing on the quantized model and adjust the pruning and quantization parameters according to the test results to ensure the stability of the model performance. For example, if it is found during the testing process that the model accuracy drops by more than 5%, then appropriately reduce the pruning ratio or increase the quantization bit width until the performance requirements are met. The finally obtained lightweight retrieval model is suitable for deployment on resource-constrained edge nodes, significantly reducing the computational overhead and storage pressure.

[0032] Edge node deployment and dynamic resource scheduling: Obtain the virtual product data requested by the user. For example, the user requests to view the rendering effect of a certain 3D furniture model through a terminal device.

[0033] Input the virtual commodity data into the lightweight retrieval model to generate the feature vectors required for rendering. The feature vectors contain key information such as the shape, material, and texture of the commodity, providing a basis for subsequent rendering.

[0034] Call the rendering engine of the edge node according to the feature vectors for real-time rendering. The edge node accelerates the rendering engine through the GPU to quickly generate high-quality 3D images.

[0035] Return the rendering result to the user terminal to complete low-latency interaction. Since the edge node is close to the user terminal, the data transmission path is significantly shortened, and the latency of rendering and interaction is effectively controlled.

[0036] The design process of the dynamic resource scheduling algorithm is as follows: Obtain the real-time load information of the edge node, including computing requirements, network latency, and node utilization. For example, computing requirements refer to the CPU and GPU resources required for the current task, network latency refers to the transmission time of data packets in the network, and node utilization refers to the occupancy ratio of the resources of the edge node. According to the real-time load information, calculate the resource allocation priority, and the formula is: Among them, represents the resource allocation priority of the edge node, is the computing requirement of the current task, is the network latency, is the node utilization, is the weight coefficient, which is used to balance various indicators. For example, when the network latency is high, increase the value to preferentially allocate resources to reduce latency.

[0037] According to the resource allocation priority, dynamically adjust the task allocation strategy of the edge node. For example, when a certain edge node is overloaded, migrate some of its tasks to the node with lower load, so as to achieve efficient resource allocation. As Figure 2 shown, this module significantly improves the overall efficiency of the system and the user experience through the feedback of real-time load information.

[0038] Implementation of the federated learning framework and the cloud-edge collaborative blockchain synchronization mechanism: On the edge node, locally train the lightweight retrieval model using local data. For example, the edge node collects user behavior data and updates the model parameters without uploading the original data, thus ensuring data privacy.

[0039] Upload the locally trained model parameters to the cloud. The cloud receives the locally trained model parameters uploaded by multiple edge nodes and aggregates them.

[0040] In the cloud, local model parameters uploaded by multiple edge nodes are aggregated to generate a global model. For example, the FedAvg algorithm is used to calculate the weighted average of the local model parameters to generate a new global model.

[0041] The global model is sent to each edge node to complete the model update.

[0042] Through the above steps, the federated learning framework realizes the continuous optimization of the model while ensuring data privacy, and solves the problem of privacy leakage in the traditional centralized training mode.

[0043] The design process of the cloud-edge collaborative blockchain synchronization mechanism is as follows: The on-chain data is divided into multiple shards, and each shard contains part of the data. For example, the transaction data is divided into several shards according to the timestamp, and each shard contains the transaction records within a certain time period.

[0044] Calculate the data transmission time after sharding, and the formula is: Among them, represents the data transmission time after sharding, N is the total data volume, is the number of shards, and P is the number of nodes participating in the consensus.

[0045] The PBFT consensus algorithm is used to verify the sharded data to ensure data consistency between the cloud and edge nodes. For example, the PBFT algorithm verifies the correctness of the sharded data through a multi-round voting mechanism and synchronizes it to all nodes. As Figure 2 shown, this module effectively reduces the communication overhead of blockchain synchronization through sharding technology and consensus algorithms, and improves the scalability and data consistency of the system.

[0046] In summary, through the collaborative action of the above-mentioned modules, the technical solution of the present invention realizes the unity of user privacy protection, accurate recommendation of virtual goods, and high-concurrency and low-latency interaction. Specifically, the homomorphic encryption technology and the zero-knowledge proof protocol ensure the security of user behavior data and transaction data; the improved deep learning retrieval model is optimized through the attention mechanism and the contrast loss function, significantly improving the parsing ability and retrieval accuracy of the multi-modal features of 3D goods; the quantization pruning process and the dynamic resource scheduling algorithm reduce the computational complexity and storage requirements of the model, while improving the overall efficiency of the system; the federated learning framework and the cloud-edge collaborative blockchain synchronization mechanism achieve the continuous optimization of the model and the consistency of data while ensuring data privacy. The technical solution of the present invention has the following technical effects: by applying the homomorphic encryption technology and the zero-knowledge proof protocol, the problem of user privacy leakage is solved; by designing and optimizing the improved deep learning retrieval model, the recommendation accuracy of virtual goods is improved; through the quantization pruning process of the lightweight retrieval model and the distributed deployment of edge nodes, the real-time rendering and low-latency interaction of virtual goods are realized; by designing the federated learning framework and the cloud-edge collaborative blockchain synchronization mechanism, the continuous optimization of the model and the consistency of data are achieved. In particular, the technical solution of the present invention provides strong technical support for the sustainable development of metaverse e-commerce.

[0047] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for blockchain encryption and deep learning retrieval in a metaverse e-commerce platform, characterized in that, Including: Adopt homomorphic encryption technology to protect the privacy of user behavior data, virtual commodity features and social network relationships, and combine the zero-knowledge proof protocol to generate tamper-proof transaction vouchers, constructing an encrypted data set containing multi-modal features; design a deep learning retrieval model, enhance the parsing ability of 3D commodity features by introducing an attention mechanism, and optimize the model parameters using a contrast loss function; perform quantization pruning on the deep learning retrieval model to obtain a lightweight retrieval model, and deploy the lightweight retrieval model to distributed edge nodes, and combine a dynamic resource scheduling algorithm to achieve real-time rendering and low-latency interaction of virtual commodities; on the edge nodes, introduce a federated learning framework to incrementally update the lightweight retrieval model, and through the way of local model training and global model aggregation, continuously optimize the model while ensuring data privacy; establish a cloud-edge collaborative blockchain synchronization mechanism, use sharding technology to reduce the amount of data transmitted on the chain, and ensure data consistency between the cloud and edge nodes through a consensus algorithm.

2. The method for blockchain encryption and deep learning retrieval for the metaverse e-commerce platform according to claim 1, wherein, The adoption of homomorphic encryption technology to protect the privacy of user behavior data, virtual commodity features and social network relationships includes: Obtain the original data of user behavior data, virtual commodity features and social network relationships; Input the original data into the homomorphic encryption module to generate an encrypted data set; Combine the zero-knowledge proof protocol to generate tamper-proof transaction vouchers, and store the transaction vouchers in the blockchain.

3. The method for blockchain encryption and deep learning retrieval used in the metaverse e-commerce platform according to claim 1, characterized in that, The process of designing the deep learning retrieval model includes: Construct a basic deep learning model, which includes an input layer, a hidden layer and an output layer; Introduce an attention mechanism module in the hidden layer to enhance the parsing ability of 3D commodity features; Define a contrast loss function to optimize the model parameters.

4. The method for blockchain encryption and deep learning retrieval used in the metaverse e-commerce platform according to claim 1, characterized in that, The process of obtaining a lightweight retrieval model by performing quantization pruning on the deep learning retrieval model includes: Obtain the trained deep learning retrieval model; Adopt model pruning technology to remove redundant parameters, and use the dynamic range quantization method to quantize floating-point weights into fixed-bit-width integers; Perform performance testing on the quantized model, and adjust the pruning and quantization parameters according to the test results to ensure the stability of the model performance.

5. The method for blockchain encryption and deep learning retrieval for a metaverse e-commerce platform according to claim 1, characterized in that, The dynamic resource scheduling algorithm includes: Obtain the real-time load information of the edge node, including computing requirements, network latency and node utilization; Calculate the resource allocation priority according to the real-time load information; Dynamically adjust the task allocation strategy of the edge node according to the resource allocation priority.

6. The method for blockchain encryption and deep learning retrieval for the metaverse e-commerce platform according to claim 1, wherein The introduction of the federated learning framework to incrementally update the lightweight retrieval model includes: On the edge node, locally train the lightweight retrieval model using local data; Upload the locally trained local model parameters to the cloud; In the cloud, aggregate the local model parameters uploaded by multiple edge nodes to generate a global model; Send the global model to each edge node to complete model update.

7. The method for blockchain encryption and deep learning retrieval used in the metaverse e-commerce platform according to claim 1, wherein, The establishment of a cloud-edge collaborative blockchain synchronization mechanism includes: Divide the data on the chain into multiple shards, and each shard contains part of the data; Calculate the data transmission time after sharding; Use a consensus algorithm to verify the sharded data and ensure data consistency between the cloud and edge nodes.

8. The method for blockchain encryption and deep learning retrieval for a metaverse e-commerce platform according to claim 1, characterized in that, The real-time rendering and low-latency interaction of the virtual goods include: Obtain the virtual goods data requested by the user; Input the virtual goods data into a lightweight retrieval model to generate the feature vectors required for rendering; According to the feature vectors, call the rendering engine of the edge node for real-time rendering; Return the rendering result to the user terminal to complete the low-latency interaction.

9. The method for blockchain encryption and deep learning retrieval for a metaverse e-commerce platform according to claim 1, wherein The zero-knowledge proof protocol generates an immutable transaction voucher, including: Obtain the transaction data and perform a hash operation on it to generate a transaction digest; Use the zero-knowledge proof algorithm to verify the transaction digest and generate a transaction voucher; Store the transaction voucher in the blockchain to ensure the immutability of the transaction data.

10. The method for blockchain encryption and deep learning retrieval for a metaverse e-commerce platform according to claim 1, wherein, The construction of an encrypted dataset containing multi-modal features includes: Obtain the original data of user behavior data, virtual goods features, and social network relationships; Preprocess the original data, including data cleaning, feature extraction, and normalization operations; Input the preprocessed data into the homomorphic encryption module to generate an encrypted multi-modal feature dataset.

Images