An unmanned aerial vehicle group security cooperation system based on blockchain and federated learning

Abstract

The present application relates to the technical field of unmanned aerial vehicle network, especially to a kind of unmanned aerial vehicle group security cooperation system based on block chain and federal learning, the present application is provided with identity management architecture module, identity registration and authentication module, communication optimization module and dynamic selection module, the identity management architecture module designs layered unmanned aerial vehicle identity management architecture, the identity registration and authentication module is registered and authenticated based on the distributed cross-domain unmanned aerial vehicle identity management mechanism, the communication optimization module responds to authentication success state, determines teacher model to update local model by knowledge distillation, the dynamic selection module is based on actual communication environment dynamic selection neighbor for the transmission process of embedding vector.The present application guarantees the security of identity management level and communication optimization level, provides protection for the whole training process, improves the computing efficiency, reduces the pressure on the center node, improves the overall communication efficiency and system response speed of unmanned aerial vehicle group.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) network technology, and in particular to a secure collaborative system for UAV swarms based on blockchain and federated learning. Background Technology

[0002] In drone network applications, identity authentication and federated learning technologies are becoming key components for ensuring system security and improving collaborative efficiency. Current identity authentication technologies mostly rely on centralized mechanisms and traditional key management systems to verify the identities of drones and their operators, ensuring that only authorized users can access the system and data. However, drone networks relying on centralized mechanisms require frequent data transmission between the central node and other nodes, which often leads to high data transmission costs and network bandwidth limitations. Federated learning, on the other hand, allows each drone to train its model locally and then send the updated results to the central node for aggregation, thereby achieving distributed learning. These technologies provide fundamental support for the collaborative operation of drone swarms.

[0003] Chinese Patent Publication No. CN119316030A discloses a method, system, device, medium, and product for secure collaboration in unmanned aerial vehicle (UAV) swarms. The method includes a UAV swarm comprising several pre-divided clusters, each cluster including a pre-selected first cluster head node and several cluster-level child nodes. The trust level of the first cluster head node meets preset conditions and is greater than that of the cluster-level child nodes. The method achieves secure communication and data sharing between UAV swarm nodes through a consensus mechanism, improving the security and reliability of the UAV swarm. By designing a hierarchical network for the UAV swarm, the UAVs are divided into several clusters, which collaborate with each other. Leveraging the hierarchical network characteristics of the UAV swarm, the consensus mechanism is divided into an intra-cluster consensus mechanism and a lightweight inter-cluster consensus mechanism, reducing the communication pressure on the cluster head and alleviating communication congestion between cluster head nodes, thus reducing communication complexity and improving the communication efficiency of the UAV swarm.

[0004] Chinese Patent Publication No. CN118301604A discloses a secure data sharing method based on blockchain-based drone collaboration. The method includes: the drone and a ground control station performing two-way authentication and key negotiation to obtain a symmetric key; the drone establishing a secure communication link with the ground control station; the drone symmetrically encrypting collected data and sending it to the ground control station through the secure communication link; the ground control station decrypting and processing the encrypted data sent by the drone, partially encrypting the data, and sending it to a cloud server, setting a data access control structure; and the cloud server encrypting the data sent by the ground control station and storing the encrypted data. This invention, based on a cloud-edge architecture, processes sensitive data locally at the edge, reducing the risk of data exposure during network transmission. By outsourcing the encryption and decryption computations to edge nodes, this invention consumes fewer resources compared to attribute-based encryption algorithms.

[0005] However, the following problems still exist in the existing technology.

[0006] Current identity authentication technologies largely rely on centralized mechanisms and traditional key management systems, which are prone to single points of failure. Sensitive data can be easily accessed by unauthorized users or devices, increasing the risk of data leakage. At the same time, current federated learning technologies rely on central servers to update and aggregate models, which can easily lead to single points of failure and performance bottlenecks, resulting in reduced model training efficiency and inconsistent results during the federated learning process. Summary of the Invention

[0007] To address this, the present invention provides a secure collaborative system for drone swarms based on blockchain and federated learning. This system solves the problems of current identity authentication technologies relying heavily on centralized mechanisms and traditional key management systems, which are prone to single points of failure and allow sensitive data to be easily accessed by unauthorized users or devices, increasing the risk of data leakage. At the same time, current federated learning technologies rely heavily on central servers to update and aggregate models, which can easily lead to single points of failure and performance bottlenecks, resulting in reduced model training efficiency and inconsistent performance during the federated learning process.

[0008] To achieve the above objectives, this invention provides a secure collaborative system for unmanned aerial vehicle (UAV) swarms based on blockchain and federated learning, comprising:

[0009] The identity management architecture module is based on a layered blockchain design to create a layered drone identity management architecture that connects various sub-trust domains.

[0010] The identity registration and authentication module, which is connected to the identity management architecture module, is designed with a distributed cross-domain drone identity management mechanism to perform drone identity registration and authentication, so as to ensure reliable identity communication between subdomain drone groups.

[0011] A communication optimization module, connected to the identity registration and authentication module, responds to a successful authentication status by determining that the teacher model performs knowledge distillation to update the local model, including...

[0012] Local training of the UAV swarm is performed. Based on the results of the local training, the embedding vector is determined. The embedding vector is filtered and aggregated. The model parameters are updated by combining the minimization of the distillation loss function and the stochastic gradient descent algorithm. Global model aggregation is then performed to obtain a converged model.

[0013] A dynamic selection module, connected to the communication optimization module, dynamically selects neighbors based on the actual communication environment during the embedding vector transmission process, including:

[0014] Calculate the user terminal distance to construct the UAV system distance matrix, determine the communication object based on the distance matrix, construct a temporary adjacency matrix, and pass the trained soft tags to neighboring UAVs.

[0015] Furthermore, the layered drone identity management architecture includes an identity management system layer, a sub-blockchain layer, a cross-domain communication layer, and a main blockchain layer.

[0016] Furthermore, the identity management architecture module connects each sub-trust domain, including,

[0017] Used to set identity attribute credentials based on the identity management system layer;

[0018] Used to store the identity attribute credentials in the sub-blockchains of each subdomain in the sub-blockchain layer;

[0019] Used to perform communication protocol conversion between domains based on the cross-domain communication layer;

[0020] This is used to enable identity authentication request communication between subdomains based on the main blockchain layer.

[0021] Furthermore, the identity registration and authentication module performs identity registration, including:

[0022] Used to generate a private key and the corresponding public key based on PUF;

[0023] Used to generate DID documents based on the identity management system that receives the public key;

[0024] Used to update the corresponding attribute accumulator in the blockchain based on the DID document and generate attribute credentials;

[0025] Used to store the attribute credentials in the memory.

[0026] Furthermore, the identity registration and authentication module performs identity authentication, including:

[0027] Used to receive authentication requests;

[0028] Used to parse the request to determine the attribute credentials that need to be verified;

[0029] This is used to determine that the attribute credential is in the blockchain and that the drone possesses the corresponding identity authentication request, which can be authenticated.

[0030] The parameters of the identity authentication request include the DID of the service and resource provider, the DID of the drone node, the access control policy, and the signature information.

[0031] Furthermore, the communication optimization module determines the teacher model, wherein,

[0032] This is used to identify drone models that are not participating in the current round of local training as teacher models.

[0033] Furthermore, the communication optimization module determines the embedding vector based on the results of the local training, including:

[0034] Used to initialize a local private model, and then trained using a shared dataset to obtain a pre-trained model;

[0035] Used to obtain local training results after training is completed;

[0036] This is used to generate embedding vectors on the shared dataset based on the local training results.

[0037] Furthermore, the communication optimization module performs filtering and aggregation operations on the embedded vector, including:

[0038] Used to filter invalid embedding vectors received by the drone;

[0039] Used to calculate similarity scores between valid embedding vectors;

[0040] This is used to select several embedding vectors with the smallest abnormal scores to form a new set;

[0041] This is used to aggregate the several embedding vectors using an aggregation function to obtain an aggregated embedding vector;

[0042] The invalid embedding vector is an embedding vector generated due to drone malfunction or malicious attack.

[0043] Furthermore, the communication optimization module updates the model parameters by combining minimizing the distillation loss function and the stochastic gradient descent algorithm, including:

[0044] Used to base on the aggregated embedding vector;

[0045] Used to determine the distillation loss function;

[0046] Used to measure the difference in probability distributions based on the distillation loss function;

[0047] This is used to combine the minimization of the distillation loss function with the stochastic gradient descent algorithm to determine the model parameters;

[0048] Used to determine the updated model based on the parameters.

[0049] Furthermore, the dynamic selection module determines the communication target based on the distance matrix, wherein,

[0050] Used to determine communication loss;

[0051] The communication object is used to determine the communication object corresponding to the minimum communication loss.

[0052] Compared with existing technologies, this invention sets up an identity management architecture module, an identity registration and authentication module, a communication optimization module, and a dynamic selection module. The identity management architecture module designs a hierarchical UAV identity management architecture. The identity registration and authentication module performs identity registration and authentication based on a distributed cross-domain UAV identity management mechanism. The communication optimization module, in response to successful authentication, determines the teacher model and performs knowledge distillation to update the local model. The dynamic selection module dynamically selects neighbors based on the actual communication environment during the embedding vector transmission process. This invention provides security for the entire training process by ensuring security at the identity management and communication optimization levels, improving computational efficiency, reducing pressure on the central node, and improving the overall communication efficiency and system response speed of the UAV swarm.

[0053] In particular, this invention identifies a method for drone node identity management and authentication based on decentralized identity, designs a layered drone identity management architecture, and uses a layered blockchain to enable interoperability between sub-trust domains. The traceability and immutability of the blockchain ensure the credibility of drone identities. In existing drone networks, training schemes lack effective identity authentication mechanisms, allowing unauthorized devices to access the network, thereby leaking privacy data or falsifying data, affecting model training. Based on this, this invention designs an identity management system layer, a sub-blockchain layer, a cross-domain communication layer, and a main blockchain layer to solve the problem of cross-domain access efficiency between different identity management systems in dynamic networks. This provides a guarantee for the identity management layer, improves computational efficiency, and reduces the pressure on the central node.

[0054] In particular, this invention combines knowledge distillation technology to perform local training on UAV swarms, determine embedding vectors, and update the model. In existing technologies, most solutions heavily rely on a central server for model aggregation and coordination. Once the central server fails, the entire system faces the risk of interruption or severe performance degradation. In addition, the computing and storage capabilities of the central server are limited, making it difficult to meet the needs of large-scale UAV networks. As the number of nodes increases, the load on the central server will continue to increase, leading to performance bottlenecks and limiting the scalability of the system. Based on this, this invention obtains a pre-trained model by training UAVs, then determines the embedding vectors, and updates the local model, providing a guarantee for communication optimization and improving the overall communication efficiency and system response speed of the UAV swarm.

[0055] In particular, by proposing a novel robust setting for dynamically selecting neighbors during information transmission, this invention addresses the challenges of decentralized federated networks. In real-world scenarios, the number of nodes in a drone network is enormous, and each node frequently needs to communicate with the central server to upload local model updates and receive the global model. This frequent communication significantly increases network overhead, especially in large-scale networks, easily leading to network congestion and latency, thus impacting training efficiency. Therefore, this invention calculates drone distances to determine communication targets, providing a safeguard for communication optimization and improving the overall communication efficiency and system response speed of the drone swarm. Attached Figure Description

[0056] Figure 1 A schematic diagram of the structure of a secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning, as an embodiment of the invention;

[0057] Figure 2 This is a schematic diagram of the hierarchical drone identity management architecture according to an embodiment of the invention;

[0058] Figure 3 This is a schematic diagram of the identity registration and authentication module in an embodiment of the invention;

[0059] Figure 4 This is a schematic diagram of the communication optimization module in an embodiment of the invention;

[0060] Figure 5 This is a schematic diagram of the structure of the dynamic selection module in an embodiment of the invention. Detailed Implementation

[0061] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0062] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0063] It should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0064] Please see Figure 1 , Figure 1 This is a schematic diagram of a secure collaborative system for drone swarms based on blockchain and federated learning, as an embodiment of the invention. The present invention provides a secure collaborative system for drone swarms based on blockchain and federated learning, comprising:

[0065] The identity management architecture module is based on a layered blockchain design to create a layered drone identity management architecture that connects various sub-trust domains.

[0066] The identity registration and authentication module, which is connected to the identity management architecture module, is designed with a distributed cross-domain drone identity management mechanism to perform drone identity registration and authentication, so as to ensure reliable identity communication between subdomain drone groups.

[0067] A communication optimization module, connected to the identity registration and authentication module, responds to a successful authentication status by determining that the teacher model performs knowledge distillation to update the local model, including...

[0068] Local training of the UAV swarm is performed. Based on the results of the local training, the embedding vector is determined. The embedding vector is filtered and aggregated. The model parameters are updated by combining the minimization of the distillation loss function and the stochastic gradient descent algorithm. Global model aggregation is then performed to obtain a converged model.

[0069] A dynamic selection module, connected to the communication optimization module, dynamically selects neighbors based on the actual communication environment during the embedding vector transmission process, including:

[0070] Calculate the user terminal distance to construct the UAV system distance matrix, determine the communication object based on the distance matrix, construct a temporary adjacency matrix, and pass the trained soft tags to neighboring UAVs.

[0071] Specifically, layered blockchain is a design pattern that divides a blockchain network into different layers, aiming to improve the scalability, performance, and security of the blockchain system. By assigning different functions and tasks to different layers, the operation of the blockchain system can be better managed and optimized.

[0072] Specifically, a teacher model is a well-trained and high-performing model used to guide the learning process of another model that is to be trained or is simpler.

[0073] Specifically, in this implementation, the steps for global model aggregation are as follows:

[0074] The drone is trained for T rounds according to the knowledge distillation mechanism. It is understood that the specific value of T is not limited. Those skilled in the art can determine it according to the actual situation, which will not be elaborated here.

[0075] Get all updated drone models;

[0076] The model is uploaded to the blockchain to complete global model aggregation.

[0077] Specifically, there are no restrictions on the specific method for constructing the distance matrix of the UAV system. For example, it can be constructed based on multidimensional scaling analysis (MDS) or on cooperative positioning. As long as the matrix can be constructed successfully, it is acceptable. Those skilled in the art can choose according to the actual situation, which will not be elaborated here.

[0078] Please see Figure 2 , Figure 2 This is a schematic diagram of the hierarchical drone identity management architecture according to an embodiment of the invention. Specifically, the hierarchical drone identity management architecture includes an identity management system layer, a sub-blockchain layer, a cross-domain communication layer, and a main blockchain layer.

[0079] Specifically, the identity management system layer consists of distributed identity management systems in each subdomain. Based on the PUF-based drone public and private key generation scheme, it calculates the accumulator corresponding to its identity attributes and generates personal attribute credentials associated with the DID document to ensure the security and uniqueness of the drone key and the authenticity and immutability of the identity attributes.

[0080] Specifically, there are no restrictions on the method of calculating the accumulator for drones. For example, a lightweight identity attribute authentication mechanism based on cryptographic accumulators can be used for calculation, which is existing technology and will not be elaborated further.

[0081] Specifically, the DID documents and accumulators in the sub-blockchain layer are stored on the sub-chain to which the sub-domain belongs, in order to ensure the integrity and immutability of the DID documents and accumulators. It is understandable that, in order to protect the privacy and security of the data on the sub-chain, the sub-blockchain can be designed as a consortium blockchain. Drone nodes are only allowed to join the sub-blockchain after authorization by the consortium blockchain.

[0082] Specifically, the cross-domain communication layer is responsible for coordinating communication between different communication domains and trusted domains, performing conversions between domains, and determining the interoperability between different trusted domains, which will not be elaborated further.

[0083] Specifically, the main blockchain layer consists of a main chain, which enables and records identity authentication request communication between subdomains, ensuring the traceability of authentication requests.

[0084] Specifically, the identity management architecture module connects various sub-trust domains, including,

[0085] Used to set identity attribute credentials based on the identity management system layer;

[0086] Used to store the identity attribute credentials in the sub-blockchains of each subdomain in the sub-blockchain layer;

[0087] Used to perform communication protocol conversion between domains based on the cross-domain communication layer;

[0088] This is used to enable identity authentication request communication between subdomains based on the main blockchain layer.

[0089] Specifically, this invention defines a method for drone node identity management and authentication based on decentralized identity, designs a layered drone identity management architecture, and uses a layered blockchain to enable interoperability between sub-trust domains. The traceability and immutability of the blockchain ensure the credibility of drone identities. In existing drone networks, training schemes lack effective identity authentication mechanisms, allowing unauthorized devices to access the network, thereby leaking privacy data or falsifying data, affecting model training. Based on this, this invention designs an identity management system layer, a sub-blockchain layer, a cross-domain communication layer, and a main blockchain layer to solve the efficiency problem of cross-domain access between different identity management systems in dynamic networks. This provides a guarantee for the identity management layer, improves computational efficiency, and reduces the pressure on the central node.

[0090] Please see Figure 3 , Figure 3 This is a schematic diagram of the identity registration and authentication module according to an embodiment of the invention. Specifically, the identity registration and authentication module performs identity registration, including:

[0091] Used to generate a private key and the corresponding public key based on PUF;

[0092] Used to generate DID documents based on the identity management system that receives the public key;

[0093] Used to update the corresponding attribute accumulator in the blockchain based on the DID document and generate attribute credentials;

[0094] Used to store the attribute credentials in the memory.

[0095] Specifically, the DID document includes attributes such as the DID number, the manufacturer of the drone, and the public key.

[0096] Understandably, to ensure the integrity and immutability of identity information, the public key is set to the public key in the DID document, and the DID document is stored on the blockchain of the subdomain, which will not be elaborated further.

[0097] Specifically, the attribute certificate needs to be encrypted using the drone's public key and sent back to the drone. Understandably, to ensure the traceability of the operation, the hash value of the stored content after this attribute update is uploaded to the main blockchain.

[0098] Specifically, the identity registration and authentication module performs identity authentication, including:

[0099] Used to receive authentication requests;

[0100] Used to parse the request to determine the attribute credentials that need to be verified;

[0101] This is used to determine that the attribute credential is in the blockchain and that the drone possesses the corresponding identity authentication request, which can be authenticated.

[0102] The parameters of the identity authentication request include the DID of the service and resource provider, the DID of the drone node, the access control policy, and the signature information.

[0103] Specifically, there are no restrictions on the specific methods for determining whether a drone possesses this attribute. For example, the Verify (ACCattcode, ACattcode) verification algorithm can be used, which is existing technology and will not be elaborated further.

[0104] Specifically, if the identity authentication request is successful, the corresponding services will be provided to the drone.

[0105] Please see Figure 4 , Figure 4 This is a schematic diagram of the communication optimization module according to an embodiment of the invention. Specifically, the communication optimization module determines the teacher model, wherein...

[0106] Used to identify drone models that did not participate in the current round of local training as teacher models;

[0107] Specifically, the communication optimization module determines the embedding vector based on the results of the local training, including:

[0108] Used to initialize a local private model, and then trained using a shared dataset to obtain a pre-trained model;

[0109] Used to obtain local training results after training is completed;

[0110] This is used to generate embedding vectors on the shared dataset based on the local training results.

[0111] Specifically, local training uses cross-entropy as the loss function, expressed by formula (1).

[0112]

[0113] In formula (1), Y represents the loss function of the entropy-dependent human-machine interface i. i Represents the private dataset D i The true label, It is a local private model M i The output, L CE This represents the cross-entropy function.

[0114] Specifically, during training, the model parameters are continuously adjusted to minimize the loss function in order to improve the model's ability to fit private data.

[0115] Specifically, the embedding vector is represented by formula (2).

[0116]

[0117] In formula (2), C represents the shared dataset. This represents the embedding generation function based on model Mi.

[0118] Specifically, the communication optimization module performs filtering and aggregation operations on the embedded vectors, including:

[0119] Used to filter invalid embedding vectors received by the drone;

[0120] Used to calculate similarity scores between valid embedding vectors;

[0121] This is used to select several embedding vectors with the smallest abnormal scores to form a new set;

[0122] This is used to aggregate the several embedding vectors using an aggregation function to obtain an aggregated embedding vector;

[0123] The invalid embedding vector is an embedding vector generated due to drone malfunction or malicious attack.

[0124] Specifically, the steps for filtering and aggregation operations are as follows.

[0125] Step S1: Calculate the Euclidean distance using formula (3).

[0126]

[0127] In formula (3), d (i,j) This represents the Euclidean distance between drone i and drone j. This represents the embedding vector of drone i. Let represent the embedding vector of drone j;

[0128] Step S2: Assign an anomaly score to each embedding vector. The anomaly score is represented by formula (4).

[0129]

[0130] In formula (4), γ i express Abnormal scores;

[0131] Step S3: Set the threshold, expressed by formula (5).

[0132]

[0133] In formula (5), ρ k Indicates the threshold;

[0134] Step S4: Select several embedding vectors with the smallest abnormal scores to form a new set Λ;

[0135] Step S5: Aggregate the embedding vectors in the vectors using an aggregation function to obtain the aggregated embedding vector, which is represented by formula (6).

[0136]

[0137] In formula (6), This represents the embedding vector of Λ in the set.

[0138] Specifically, this invention combines knowledge distillation technology to perform local training on a drone swarm, determine embedding vectors, and update the model. In existing technologies, most solutions heavily rely on a central server for model aggregation and coordination. If the central server fails, the entire system faces the risk of interruption or severe performance degradation. Furthermore, the computing and storage capabilities of the central server are limited, making it difficult to meet the needs of large-scale drone networks. As the number of nodes increases, the load on the central server will continuously increase, leading to performance bottlenecks and limiting the system's scalability. Based on this, this invention obtains a pre-trained model by training drones, then determines the embedding vectors, and updates the local model, providing a guarantee for communication optimization and improving the overall communication efficiency and system response speed of the drone swarm.

[0139] Specifically, the communication optimization module updates model parameters by combining minimizing the distillation loss function with the stochastic gradient descent algorithm, including:

[0140] Used to base on the aggregated embedding vector;

[0141] Used to determine the distillation loss function;

[0142] Used to measure the difference in probability distributions based on the distillation loss function;

[0143] This is used to combine the minimization of the distillation loss function with the stochastic gradient descent algorithm to determine the model parameters;

[0144] Used to determine the updated model based on the parameters.

[0145] Specifically, the distillation loss function is expressed by formula (7).

[0146]

[0147] In formula (7), D KL V represents the Kullback-Leibler divergence. k Represents the aggregated embedding vector. Let represent the local embedding vector, which is expressed by formula (8).

[0148]

[0149] Specifically, the Kullback-Leibler divergence is used to measure the difference between two probability distributions.

[0150] Specifically, the steps of the stochastic gradient descent algorithm are as follows.

[0151] Initialize parameters, including randomly initializing model parameters, learning rate, and setting maximum number of iterations or convergence conditions;

[0152] In each iteration, a sample is randomly selected;

[0153] Calculate the gradient of the loss function based on the selected samples;

[0154] Based on the calculated gradient, update the model parameters to reduce the value of the loss function;

[0155] Repeat the above steps until the stopping condition is met.

[0156] Specifically, the stopping conditions could be reaching the maximum number of iterations, the change in the loss function being less than a certain threshold, or the performance on the validation set no longer improving, which will not be elaborated further.

[0157] Specifically, after T rounds of training according to the knowledge distillation mechanism, the drones upload their updated models to the blockchain for global model aggregation, resulting in a converged model. This converged model is the result of collaborative training by several drones, combining the local models of several drones, and has strong adaptability and robustness, thus improving the collaborative efficiency and learning outcomes of the drones.

[0158] Please see Figure 5 , Figure 5 This is a schematic diagram of the dynamic selection module according to an embodiment of the invention. Specifically, the dynamic selection module determines the communication target based on the distance matrix, wherein...

[0159] Used to determine communication loss;

[0160] The communication object is used to determine the communication object corresponding to the minimum communication loss.

[0161] Specifically, this invention proposes a novel robust setting for dynamically selecting neighbors during information transmission. In real-world decentralized federation scenarios, the number of nodes in a drone network is enormous, and each node needs to frequently communicate with the central server to upload local model updates and receive the global model. This frequent communication significantly increases network overhead, especially in large-scale networks, easily leading to network congestion and latency, thus affecting training efficiency. Therefore, this invention calculates drone distances to determine communication targets, providing a safeguard for communication optimization and improving the overall communication efficiency and system response speed of the drone swarm.

[0162] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0163] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A secure collaborative system for unmanned aerial vehicle (UAV) swarms based on blockchain and federated learning, characterized in that, include: The identity management architecture module is based on a layered blockchain design to create a layered drone identity management architecture that connects various sub-trust domains. The identity registration and authentication module, which is connected to the identity management architecture module, is designed with a distributed cross-domain drone identity management mechanism to perform drone identity registration and authentication, so as to ensure reliable identity communication between subdomain drone groups. The communication optimization module, connected to the identity registration and authentication module, responds to a successful authentication status by determining that the teacher model should undergo knowledge distillation to update the local model. include, Local training of the UAV swarm is performed. Based on the results of the local training, the embedding vector is determined. The embedding vector is filtered and aggregated. The model parameters are updated by combining the minimization of the distillation loss function and the stochastic gradient descent algorithm. Global model aggregation is then performed to obtain a converged model. A dynamic selection module, connected to the communication optimization module, dynamically selects neighbors based on the actual communication environment during the embedding vector transmission process. include, Calculate the user terminal distance to construct the UAV system distance matrix, determine the communication object based on the distance matrix, construct a temporary adjacency matrix, and pass the trained soft tags to neighboring UAVs.

2. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 1, characterized in that, The hierarchical drone identity management architecture includes an identity management system layer, a sub-blockchain layer, a cross-domain communication layer, and a main blockchain layer.

3. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 2, characterized in that, The identity management architecture module connects various sub-trust domains, including: Used to set identity attribute credentials based on the identity management system layer; Used to store the identity attribute credentials in the sub-blockchains of each subdomain in the sub-blockchain layer; Used to perform communication protocol conversion between domains based on the cross-domain communication layer; This is used to enable identity authentication request communication between subdomains based on the main blockchain layer.

4. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 1, characterized in that, The identity registration and authentication module performs identity registration. include, Used to generate a private key and the corresponding public key based on PUF; Used to generate DID documents based on the identity management system that receives the public key; Used to update the corresponding attribute accumulator in the blockchain based on the DID document and generate attribute credentials; Used to store the attribute credentials in the memory.

5. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 4, characterized in that, The identity registration and authentication module performs identity authentication, including: Used to receive authentication requests; Used to parse the request to determine the attribute credentials that need to be verified; This is used to determine that the attribute credential is in the blockchain and that the drone possesses the corresponding identity authentication request, which can be authenticated. The parameters of the identity authentication request include the DID of the service and resource provider, the DID of the drone node, the access control policy, and the signature information.

6. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 1, characterized in that, The communication optimization module determines the teacher model, wherein... This is used to identify drone models that are not participating in the current round of local training as teacher models.

7. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 1, characterized in that, The communication optimization module determines the embedding vector based on the results of the local training, including: Used to initialize a local private model, and then trained using a shared dataset to obtain a pre-trained model; Used to obtain local training results after training is completed; This is used to generate embedding vectors on the shared dataset based on the local training results.

8. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 1, characterized in that, The communication optimization module performs filtering and aggregation operations on the embedded vectors, including: Used to filter invalid embedding vectors received by the drone; Used to calculate similarity scores between valid embedding vectors; This is used to select several embedding vectors with the smallest abnormal scores to form a new set; This is used to aggregate the several embedding vectors using an aggregation function to obtain an aggregated embedding vector; The invalid embedding vector is an embedding vector generated due to drone malfunction or malicious attack.

9. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 8, characterized in that, The communication optimization module updates model parameters by combining minimizing the distillation loss function and the stochastic gradient descent algorithm, including: Used to base on the aggregated embedding vector; Used to determine the distillation loss function; Used to measure the difference in probability distributions based on the distillation loss function; This is used to combine the minimization of the distillation loss function with the stochastic gradient descent algorithm to determine the model parameters; Used to determine the updated model based on the parameters.

10. The secure collaborative system for unmanned aerial vehicle swarms based on blockchain and federated learning according to claim 1, characterized in that, The dynamic selection module determines the communication target based on the distance matrix, wherein, Used to determine communication loss; The communication object is used to determine the communication object corresponding to the minimum communication loss.

Images