Federal learning-based emergency rescue data privacy protection and collaborative analysis system

Abstract

The invention discloses an emergency rescue data privacy protection and collaborative analysis system based on federated learning, and belongs to the technical field of emergency rescue data privacy protection and collaborative analysis. A dynamic trust evaluation module; the federated learning privacy protection engine module is used for realizing cross-domain data collaborative training by adopting a secure multi-party computing model; and the block chain intelligent contract module is used for deploying a data access control strategy and a credible auditing rule. According to the method, a global trusted environment is established through an end-side cloud three-level trust chain, the reliability of equipment is quantified through dynamic trust evaluation, data collaboration of privacy protection is realized through federated learning, an auditing strategy is automatically executed through an intelligent contract, an efficient encryption technology for guaranteeing safety and effectively reducing resource consumption is adopted, and the urban emergency data safety problem is solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of emergency rescue data privacy protection and collaborative analysis, and particularly relates to an emergency rescue data privacy protection and collaborative analysis system based on federated learning. BACKGROUND

[0002] At present, various types of devices are widely used in urban emergency management, covering public health, disaster warning, traffic control and other fields. The technical characteristics, working principles and safety requirements of each device are different, which makes it difficult for a single security mechanism to meet the safety requirements of all devices. In the process of sending and receiving emergency data, if unauthenticated devices access the system, it may bring great security risks such as data leakage, identity forgery, service denial and other problems, further exacerbating the vulnerability of urban management systems. Therefore, how to accurately authenticate, encrypt and dynamically monitor a large number of heterogeneous emergency devices to ensure the safe and reliable operation of urban emergency devices has become a key to ensuring the safe and reliable transmission of urban emergency data. With the rapid development of artificial intelligence and big data analysis technology, attackers can use complex network attack technology to implement highly concealed attacks on urban emergency systems, which puts higher security requirements on existing emergency data transmission. At present, most emergency devices have limited computing power and resources, and existing security protection mechanisms often bring high computational overhead, making it difficult to balance security and efficiency. Especially in high-load and low-resource emergency environments, existing security transmission mechanisms may cause devices to be unable to bear due to excessive consumption of computing resources, thereby affecting the real-time and accuracy of data and endangering the reliability of emergency decision-making. Therefore, an emergency rescue data privacy protection and collaborative analysis system based on federated learning is needed, which can not only guarantee security but also effectively reduce resource consumption of efficient encryption technology, and become a solution to urban emergency data security problems. SUMMARY

[0003] The purpose of the present application is to overcome the shortcomings of the prior art and provide an emergency rescue data privacy protection and collaborative analysis system based on federated learning, which solves the problems in the background art.

[0004] The objective of this invention is achieved as follows: An emergency rescue data privacy protection and collaborative analysis system based on federated learning, comprising: a trusted computing architecture module for edge-cloud collaboration: establishing a three-tiered trust chain among terminal devices, edge nodes, and the cloud center through physical roots of trust; a dynamic trust assessment module: integrating spatiotemporal variables and behavioral metrics to calculate the dynamic trustworthiness of devices, where the spatiotemporal variables include timestamps, geographical locations, and environmental parameters, and the behavioral metrics include response time, error rate, and collaborative efficiency; a federated learning privacy protection engine module: employing a secure multi-party computation model to achieve cross-domain data collaborative training; and a blockchain smart contract module: used to deploy data access control policies and trusted audit rules. The working principle is as follows: establishing a globally trusted environment through the three-tiered trust chain of edge-cloud, dynamic trust assessment quantifies device reliability, federated learning achieves privacy-protected data collaboration, and smart contracts automatically execute audit policies. In use, the terminal activates the TPM chip to verify the firmware signature, the edge gateway collects device response time / error rate data every 5 seconds, the cloud integrates spatiotemporal data to calculate dynamic trustworthiness, and the smart contract triggers abnormal data isolation. Traditional emergency response systems suffer from trust gaps, data silos, and inefficient manual auditing. Highly efficient encryption technologies that ensure security and effectively reduce resource consumption address the issue of urban emergency data security.

[0005] Furthermore, the dynamic trust assessment module includes:

[0006] Spatiotemporal credibility calculation unit: through formula,

[0007]

[0008] The impact of environmental factors is quantified; in the formula, E(t,s) is the spatiotemporal environmental factor function; t is the timestamp; s is the spatial coordinate; E i (t,s) represents the value of the i-th type of environmental parameter at the spatiotemporal point (t,s); n is the total number of environmental parameter types; ω i For the i-th type of environmental parameter, the weight coefficient is denoted as .

[0009] Behavioral measurement acquisition unit: used for real-time monitoring of device response time, error rate, and collaboration capability indicators;

[0010] Credibility dynamic correction unit: updates credibility according to time windows.

[0011] The working principle is as follows: the spatiotemporal environmental factor function integrates the dynamic impact of disaster environment, and the weights are adaptively adjusted according to the disaster type to drive accurate calculation of credibility. When used, in the earthquake scenario, the network latency weight ω1 = 0.6 is set, and in the flood scenario, the meteorological level weight ω2 = 0.7 is set. The environmental risk value is calculated in real time for each kilometer grid, which solves the problem of misjudgment caused by the static trust model ignoring spatiotemporal changes.

[0012] Furthermore, the behavior measurement acquisition unit is used to define a response time threshold of no more than 1 second, a malicious data identification accuracy of ≥80%, and to calculate behavior credibility. Through the dual constraints of the response time threshold (≤1 second) and malicious data identification (≥80% accuracy), the reliability of behavior credibility calculation is improved, avoiding evaluation bias caused by the lack of quantitative standards in behavior monitoring.

[0013] Furthermore, the federated learning privacy protection engine module includes the implementation of: intra-domain data training: each participant trains a sub-model locally; gradient secure aggregation: encrypts model parameters based on homomorphic encryption technology;

[0014] Dynamic weight update mechanism: The weights of the federated model are adjusted based on the value of the data. The update formula is as follows:

[0015]

[0016] In the formula, ω i T represents the weighting coefficient of the i-th type of environmental parameter; exposure is the data exposure time; i is the index of federated learning participants.

[0017] It enables the quantification of the value of time-sensitive data, and homomorphic encryption ensures the security of gradient aggregation, avoiding the dilution of the value of time-sensitive data caused by the equal weighting of traditional federated learning.

[0018] Furthermore, the blockchain smart contract module includes: a data lifecycle auditing unit for recording the entire data process from acquisition to destruction; a self-destruct key mechanism for automatically destroying the key when unauthorized access is detected; and a dynamic permission adjustment unit for allocating data access permissions based on the device's trust level. Trust level drives dynamic permission adjustment, and the self-destruct key mechanism enables second-level response to unauthorized access.

[0019] The implementation process is as follows: / / Smart contract pseudocode if device trustworthiness < 0.6:

[0020] revokeAccess(); / / Revoke permissions if malicious attack detected:

[0021] selfDestructKey(); / / Self-destruct key

[0022] Addressing security vulnerabilities caused by the lack of adaptability in fixed permission policies.

[0023] Furthermore, it also includes a trusted group construction unit, which dynamically generates a task-driven device cluster based on device collaboration trustworthiness. The overall trustworthiness calculation formula for the cluster is:

[0024]

[0025] In the formula, T group The overall trustworthiness of the device cluster; M is the total number of devices in the cluster; k is the index number of the device within the cluster; φ represents the dynamic reliability of the k-th device. k This represents the weight coefficient for the k-th device. The weight amplifies the impact of critical devices, and cluster reliability mitigates the risk of single-point failures. The implementation process is as follows:

[0026] Equipment weight = [Command Terminal: 1.5, Life Detector: 1.2, Camera: 1.0]

[0027] Cluster reliability = (1.5×0.9+1.2×0.85+1.0×0.7) / (1.5+1.2+1.0)≈0.84, avoiding the defect of the equal cluster strategy ignoring the critical differences of devices.

[0028] Furthermore, it also includes: a lightweight cryptographic module: employing the national cryptographic algorithm SM4-GCM for data encryption; and a key security management unit: generating and storing session keys within a trusted execution environment (TEE). The lightweight cryptographic module includes implementations of: generating key pairs using the SM2 algorithm (4096 bits in length); calculating data hash values ​​using SM3; and encrypting the data stream using SM4-CTR mode. SM4-CTR stream encryption combined with SM3 hash verification, and key management within the TEE to resist physical attacks, is implemented as follows:

[0029] sm4_ctr_encrypt(data, 4096_key); / / Encryption in CTR mode

[0030] sm3_hash = hash(data); / / Generate authentication tags.

[0031] This addresses the resource bottleneck issue of traditional encryption in high-load emergency scenarios.

[0032] Furthermore, it also includes an emergency resource optimization and scheduling module, which achieves resource allocation through an improved differential evolution algorithm, specifically including:

[0033] Establish a resource scheduling model:

[0034]

[0035] In the formula, V represents the emergency resource scheduling priority score; N represents the total number of resource types participating in the scheduling; CVSS i Score the vulnerability threat of resource type i; ω i λ is the weighting coefficient for the i-th type of resource; λ is the time decay coefficient; T i Let be the idle time of the i-th type of resource; e≈2.718 is the natural constant.

[0036] Visualized decision support: Based on GIS maps, resource distribution and scheduling schemes are displayed.

[0037] The working principle is as follows: CVSS vulnerability-based scheduling prioritizes high-risk resources, and a time decay factor penalizes idle resources, thus addressing the shortcomings of scheduling models that ignore security vulnerabilities and time costs.

[0038] Furthermore, the terminal device includes a trusted box hardware deployment, which comprises: a Raspberry Pi main control platform; a Chinese national cryptographic algorithm cryptographic chip; a side-channel protection circuit: resisting differential power consumption analysis through dynamic noise injection; and a trusted boot mechanism: verifying the hash values ​​of the Bootloader, kernel, and application layer step by step. The side-channel protection circuit addresses the problem of weak physical security protection of equipment at disaster sites by using power supply noise injection interference power consumption analysis, trusted boot chain verification of firmware integrity, and dynamic noise injection power path.

[0039] The beneficial effects of this invention are as follows: A globally trusted environment is established through a three-tiered trust chain (edge, cloud, and terminal); dynamic trust assessment quantifies device reliability; federated learning enables privacy-preserving data collaboration; and smart contracts automate audit strategies. In use, the terminal activates the TPM chip to verify the firmware signature; the edge gateway collects device response time / error rate data every 5 seconds; the cloud integrates spatiotemporal data to calculate dynamic trustworthiness; and smart contracts trigger abnormal data isolation. This invention addresses the trust fragmentation, data silos, and inefficiency of manual auditing inherent in traditional emergency rescue systems. It employs a highly efficient encryption technology that ensures security and effectively reduces resource consumption, solving urban emergency data security issues. The spatiotemporal environmental factor function comprehensively considers the dynamic impact of the disaster environment, with weights adaptively adjusted according to disaster type, driving accurate trustworthiness calculation. In use, for earthquake scenarios, the network latency weight ω1 = 0.6; for flood scenarios, the meteorological level weight ω2 = 0.7. Environmental risk values ​​are calculated in real-time per kilometer grid, resolving misjudgments caused by static trust models ignoring spatiotemporal changes. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0041] The present invention will now be described in further detail with reference to the accompanying drawings. It should be noted that this is only for the purpose of more clearly illustrating and explaining the present invention.

[0042] Example 1

[0043] like Figure 1As shown, this embodiment discloses an emergency rescue data privacy protection and collaborative analysis system based on federated learning, which includes: a trusted computing architecture module for edge-cloud collaboration: establishing a three-level trust chain between terminal devices, edge nodes, and the cloud center through physical roots of trust; a dynamic trust assessment module: integrating spatiotemporal variables and behavioral metrics to calculate the dynamic trustworthiness of devices, where the spatiotemporal variables include timestamps, geographical locations, and environmental parameters, and the behavioral metrics include response time, error rate, and collaborative efficiency; a federated learning privacy protection engine module: employing a secure multi-party computation model to achieve cross-domain data collaborative training; and a blockchain smart contract module: used to deploy data access control policies and trusted audit rules. The working principle is as follows: establishing a fully trusted environment through the three-level trust chain between the edge and cloud, quantifying device reliability through dynamic trust assessment, achieving privacy-protected data collaboration through federated learning, and automatically executing audit policies through smart contracts. In use, the terminal activates the TPM chip to verify the firmware signature, the edge gateway collects device response time / error rate every 5 seconds, the cloud integrates spatiotemporal data to calculate dynamic trustworthiness, and the smart contract triggers abnormal data isolation. Traditional emergency response systems suffer from trust gaps, data silos, and inefficient manual auditing. Highly efficient encryption technologies that ensure security and effectively reduce resource consumption address the issue of urban emergency data security.

[0044] Example 2

[0045] like Figure 1 As shown, this embodiment discloses an emergency rescue data privacy protection and collaborative analysis system based on federated learning, which includes: a trusted computing architecture module for edge-cloud collaboration: establishing a three-level trust chain between terminal devices, edge nodes, and the cloud center through physical roots of trust; a dynamic trust assessment module: integrating spatiotemporal variables and behavioral metrics to calculate the dynamic trustworthiness of devices, where the spatiotemporal variables include timestamps, geographical locations, and environmental parameters, and the behavioral metrics include response time, error rate, and collaborative efficiency; a federated learning privacy protection engine module: employing a secure multi-party computation model to achieve cross-domain data collaborative training; and a blockchain smart contract module: used to deploy data access control policies and trusted audit rules. The working principle is as follows: establishing a fully trusted environment through the three-level trust chain between the edge and cloud, quantifying device reliability through dynamic trust assessment, achieving privacy-protected data collaboration through federated learning, and automatically executing audit policies through smart contracts. In use, the terminal activates the TPM chip to verify the firmware signature, the edge gateway collects device response time / error rate every 5 seconds, the cloud integrates spatiotemporal data to calculate dynamic trustworthiness, and the smart contract triggers abnormal data isolation. Traditional emergency response systems suffer from trust gaps, data silos, and inefficient manual auditing. Highly efficient encryption technologies that ensure security and effectively reduce resource consumption address the issue of urban emergency data security.

[0046] For better results, the dynamic trust assessment module includes:

[0047] Spatiotemporal credibility calculation unit: through formula,

[0048]

[0049] The impact of environmental factors is quantified; in the formula, E(t,s) is the spatiotemporal environmental factor function; t is the timestamp; s is the spatial coordinate; E i (t,s) represents the value of the i-th type of environmental parameter at the spatiotemporal point (t,s); n is the total number of environmental parameter types; ω i For the i-th type of environmental parameter, the weight coefficient is denoted as .

[0050] Behavioral measurement acquisition unit: used for real-time monitoring of device response time, error rate, and collaboration capability indicators;

[0051] Credibility dynamic correction unit: updates credibility according to time windows.

[0052] The working principle is as follows: the spatiotemporal environmental factor function integrates the dynamic impact of disaster environment, and the weights are adaptively adjusted according to the disaster type to drive accurate calculation of credibility. When used, in the earthquake scenario, the network latency weight ω1 = 0.6 is set, and in the flood scenario, the meteorological level weight ω2 = 0.7 is set. The environmental risk value is calculated in real time for each kilometer grid, which solves the problem of misjudgment caused by the static trust model ignoring spatiotemporal changes.

[0053] To achieve better results, the behavior measurement acquisition unit defines a response time threshold of no more than 1 second, a malicious data identification accuracy of ≥80%, and calculates behavior credibility. By using the dual constraints of the response time threshold (≤1 second) and malicious data identification (≥80% accuracy), the reliability of behavior credibility calculation is improved, avoiding evaluation bias caused by the lack of quantitative standards in behavior monitoring.

[0054] To achieve better results, the federated learning privacy protection engine module includes the following implementations: intra-domain data training: each participant trains its sub-model locally; gradient secure aggregation: model parameters are encrypted based on homomorphic encryption technology;

[0055] Dynamic weight update mechanism: The weights of the federated model are adjusted based on the value of the data. The update formula is as follows:

[0056]

[0057] In the formula, ω i T represents the weighting coefficient of the i-th type of environmental parameter; exposure is the data exposure time; i is the index of federated learning participants.

[0058] It enables the quantification of the value of time-sensitive data, and homomorphic encryption ensures the security of gradient aggregation, avoiding the dilution of the value of time-sensitive data caused by the equal weighting of traditional federated learning.

[0059] For better performance, the blockchain smart contract module includes: a data lifecycle auditing unit for recording the entire data process from acquisition to destruction; a self-destructing key mechanism for automatically destroying the key when unauthorized access is detected; and a dynamic permission adjustment unit for allocating data access permissions based on the device's trust level. Trust level drives dynamic permission adjustment, and the self-destructing key mechanism enables second-level response to unauthorized access.

[0060] The implementation process is as follows: / / Smart contract pseudocode if device trustworthiness < 0.6:

[0061] revokeAccess(); / / Revoke permissions if malicious attack detected:

[0062] selfDestructKey(); / / Self-destruct key

[0063] Addressing security vulnerabilities caused by the lack of adaptability in fixed permission policies.

[0064] For better results, a trusted group building unit is also included. This unit dynamically generates a task-driven device cluster based on device collaboration trustworthiness. The overall trustworthiness of the cluster is calculated using the following formula:

[0065]

[0066] In the formula, T group The overall trustworthiness of the device cluster; M is the total number of devices in the cluster; k is the index number of the device within the cluster; φ represents the dynamic reliability of the k-th device. k This represents the weight coefficient for the k-th device. The weight amplifies the impact of critical devices, and cluster reliability mitigates the risk of single-point failures. The implementation process is as follows:

[0067] Equipment weight = [Command Terminal: 1.5, Life Detector: 1.2, Camera: 1.0]

[0068] Cluster reliability = (1.5×0.9+1.2×0.85+1.0×0.7) / (1.5+1.2+1.0)≈0.84, avoiding the defect of the equal cluster strategy ignoring the critical differences of devices.

[0069] For better performance, the system also includes: a lightweight cryptographic module: employing the national cryptographic algorithm SM4-GCM for data encryption; and a key security management unit: generating and storing session keys within a trusted execution environment (TEE). The lightweight cryptographic module includes implementations of: generating key pairs using the SM2 algorithm (4096 bits in length); calculating data hash values ​​using SM3; and encrypting the data stream using SM4-CTR mode. SM4-CTR stream encryption, combined with SM3 hash verification and key management within the TEE, protects against physical attacks. The implementation process is as follows:

[0070] sm4_ctr_encrypt(data, 4096_key); / / Encryption in CTR mode

[0071] sm3_hash = hash(data); / / Generate authentication tags.

[0072] This addresses the resource bottleneck issue of traditional encryption in high-load emergency scenarios.

[0073] To achieve better results, an emergency resource optimization and scheduling module is also included. This module uses an improved differential evolution algorithm to allocate resources, specifically including:

[0074] Establish a resource scheduling model:

[0075]

[0076] In the formula, V represents the emergency resource scheduling priority score; N represents the total number of resource types participating in the scheduling; CVSS i Score the vulnerability threat of resource type i; ω i λ is the weighting coefficient for the i-th type of resource; λ is the time decay coefficient; T i Let be the idle time of the i-th type of resource; e≈2.718 is the natural constant.

[0077] Visualized decision support: Based on GIS maps, resource distribution and scheduling schemes are displayed.

[0078] The working principle is as follows: CVSS vulnerability-based scheduling prioritizes high-risk resources, and a time decay factor penalizes idle resources, thus addressing the shortcomings of scheduling models that ignore security vulnerabilities and time costs.

[0079] Example 3

[0080] like Figure 1As shown, this embodiment discloses an emergency rescue data privacy protection and collaborative analysis system based on federated learning, which includes: a trusted computing architecture module for edge-cloud collaboration: establishing a three-level trust chain between terminal devices, edge nodes, and the cloud center through physical roots of trust; a dynamic trust assessment module: integrating spatiotemporal variables and behavioral metrics to calculate the dynamic trustworthiness of devices, where the spatiotemporal variables include timestamps, geographical locations, and environmental parameters, and the behavioral metrics include response time, error rate, and collaborative efficiency; a federated learning privacy protection engine module: employing a secure multi-party computation model to achieve cross-domain data collaborative training; and a blockchain smart contract module: used to deploy data access control policies and trusted audit rules. The working principle is as follows: establishing a fully trusted environment through the three-level trust chain between the edge and cloud, quantifying device reliability through dynamic trust assessment, achieving privacy-protected data collaboration through federated learning, and automatically executing audit policies through smart contracts. In use, the terminal activates the TPM chip to verify the firmware signature, the edge gateway collects device response time / error rate every 5 seconds, the cloud integrates spatiotemporal data to calculate dynamic trustworthiness, and the smart contract triggers abnormal data isolation. Traditional emergency response systems suffer from trust gaps, data silos, and inefficient manual auditing. Highly efficient encryption technologies that ensure security and effectively reduce resource consumption address the issue of urban emergency data security.

[0081] For better results, the dynamic trust assessment module includes:

[0082] Spatiotemporal credibility calculation unit: through formula,

[0083]

[0084] The impact of environmental factors is quantified; in the formula, E(t,s) is the spatiotemporal environmental factor function; t is the timestamp; s is the spatial coordinate; E i (t,s) represents the value of the i-th type of environmental parameter at the spatiotemporal point (t,s); n is the total number of environmental parameter types; ω i For the i-th type of environmental parameter, the weight coefficient is denoted as .

[0085] Behavioral measurement acquisition unit: used for real-time monitoring of device response time, error rate, and collaboration capability indicators;

[0086] Credibility dynamic correction unit: updates credibility according to time windows.

[0087] The working principle is as follows: the spatiotemporal environmental factor function integrates the dynamic impact of disaster environment, and the weights are adaptively adjusted according to the disaster type to drive accurate calculation of credibility. When used, in the earthquake scenario, the network latency weight ω1 = 0.6 is set, and in the flood scenario, the meteorological level weight ω2 = 0.7 is set. The environmental risk value is calculated in real time for each kilometer grid, which solves the problem of misjudgment caused by the static trust model ignoring spatiotemporal changes.

[0088] To achieve better results, the behavior measurement acquisition unit defines a response time threshold of no more than 1 second, a malicious data identification accuracy of ≥80%, and calculates behavior credibility. By using the dual constraints of the response time threshold (≤1 second) and malicious data identification (≥80% accuracy), the reliability of behavior credibility calculation is improved, avoiding evaluation bias caused by the lack of quantitative standards in behavior monitoring.

[0089] To achieve better results, the federated learning privacy protection engine module includes the following implementations: intra-domain data training: each participant trains its sub-model locally; gradient secure aggregation: model parameters are encrypted based on homomorphic encryption technology;

[0090] Dynamic weight update mechanism: The weights of the federated model are adjusted based on the value of the data. The update formula is as follows:

[0091]

[0092] In the formula, ω i T represents the weighting coefficient of the i-th type of environmental parameter; exposure is the data exposure time; i is the index of federated learning participants.

[0093] It enables the quantification of the value of time-sensitive data, and homomorphic encryption ensures the security of gradient aggregation, avoiding the dilution of the value of time-sensitive data caused by the equal weighting of traditional federated learning.

[0094] For better performance, the blockchain smart contract module includes: a data lifecycle auditing unit for recording the entire data process from acquisition to destruction; a self-destructing key mechanism for automatically destroying the key when unauthorized access is detected; and a dynamic permission adjustment unit for allocating data access permissions based on the device's trust level. Trust level drives dynamic permission adjustment, and the self-destructing key mechanism enables second-level response to unauthorized access.

[0095] The implementation process is as follows: / / Smart contract pseudocode if device trustworthiness < 0.6:

[0096] revokeAccess(); / / Revoke permissions if malicious attack detected:

[0097] selfDestructKey(); / / Self-destruct key

[0098] Addressing security vulnerabilities caused by the lack of adaptability in fixed permission policies.

[0099] For better results, a trusted group building unit is also included. This unit dynamically generates a task-driven device cluster based on device collaboration trustworthiness. The overall trustworthiness of the cluster is calculated using the following formula:

[0100]

[0101] In the formula, T group The overall trustworthiness of the device cluster; M is the total number of devices in the cluster; k is the index number of the device within the cluster; φ represents the dynamic reliability of the k-th device. k This represents the weight coefficient for the k-th device. The weight amplifies the impact of critical devices, and cluster reliability mitigates the risk of single-point failures. The implementation process is as follows:

[0102] Equipment weight = [Command Terminal: 1.5, Life Detector: 1.2, Camera: 1.0]

[0103] Cluster reliability = (1.5×0.9+1.2×0.85+1.0×0.7) / (1.5+1.2+1.0)≈0.84, avoiding the defect of the equal cluster strategy ignoring the critical differences of devices.

[0104] For better performance, the system also includes: a lightweight cryptographic module: employing the national cryptographic algorithm SM4-GCM for data encryption; and a key security management unit: generating and storing session keys within a trusted execution environment (TEE). The lightweight cryptographic module includes implementations of: generating key pairs using the SM2 algorithm (4096 bits in length); calculating data hash values ​​using SM3; and encrypting the data stream using SM4-CTR mode. SM4-CTR stream encryption, combined with SM3 hash verification and key management within the TEE, protects against physical attacks. The implementation process is as follows:

[0105] sm4_ctr_encrypt(data, 4096_key); / / Encryption in CTR mode

[0106] sm3_hash = hash(data); / / Generate authentication tags.

[0107] This addresses the resource bottleneck issue of traditional encryption in high-load emergency scenarios.

[0108] To achieve better results, an emergency resource optimization and scheduling module is also included. This module uses an improved differential evolution algorithm to allocate resources, specifically including:

[0109] Establish a resource scheduling model:

[0110]

[0111] In the formula, V represents the emergency resource scheduling priority score; N represents the total number of resource types participating in the scheduling; CVSS i Score the vulnerability threat of resource type i; ω i λ is the weighting coefficient for the i-th type of resource; λ is the time decay coefficient; T i Let be the idle time of the i-th type of resource; e≈2.718 is the natural constant.

[0112] Visualized decision support: Based on GIS maps, resource distribution and scheduling schemes are displayed.

[0113] The working principle is as follows: CVSS vulnerability-based scheduling prioritizes high-risk resources, and a time decay factor penalizes idle resources, thus addressing the shortcomings of scheduling models that ignore security vulnerabilities and time costs.

[0114] For better performance, the terminal device includes a trusted box hardware deployment, which comprises: a Raspberry Pi main control platform; a Chinese national cryptographic algorithm cryptographic chip; a side-channel protection circuit: resisting differential power consumption analysis through dynamic noise injection; and a trusted boot mechanism: verifying the hash values ​​of the bootloader, kernel, and application layer step by step. The side-channel protection circuit addresses the weakness of physical security protection for equipment at disaster sites by using power supply noise injection interference power consumption analysis, a trusted boot chain to verify firmware integrity, and dynamic noise injection power path.

[0115] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An emergency rescue data privacy protection and collaborative analysis system based on federated learning, characterized in that, Comprise: End-edge-cloud collaborative trusted computing architecture module: a three-level trust chain of terminal device, edge node and cloud center is established through physical trusted root; Dynamic trust evaluation module: the dynamic trust degree of the device is calculated by fusing space-time variables and behavior metrics, the space-time variables include timestamp, geographic location and environmental parameters, and the behavior metrics include response time, error rate and cooperation efficiency; Federal learning privacy protection engine module: cross-domain data collaborative training is realized by using secure multi-party computation model; Blockchain smart contract module: used for deploying data access control strategy and trusted audit rules.

2. The emergency rescue data privacy protection and collaborative analysis system based on federated learning according to claim 1, characterized in that, The dynamic trust evaluation module comprises: Space-time trust degree calculation unit: through formula, quantifying the impact of environmental factors; where E(t,s) is the spatio-temporal environmental factor function; t is the time stamp; s is the spatial coordinate; E i (t,s) the value of the ith type of environmental parameter at the spatio-temporal point (t,s); n is the total number of environmental parameter types; ω i is the weight coefficient of the ith type of environmental parameter; Behavior metric acquisition unit: used for real-time monitoring of device response time, error rate and cooperation capability indicators; Trust degree dynamic correction unit: update the trust degree according to the time window.

3. The emergency rescue data privacy protection and collaborative analysis system based on federated learning according to claim 2, characterized in that, The behavior metric acquisition unit is used to define that the response time threshold is not more than 1 second, the accuracy of malicious data identification is greater than or equal to 80%, and the behavior trust degree is calculated.

4. The emergency rescue data privacy protection and collaborative analysis system based on federated learning according to claim 1, characterized in that, The federal learning privacy protection engine module comprises: Intra-domain data training: each participant trains a sub-model locally; Gradient secure aggregation: encrypt model parameters based on homomorphic encryption technology; Dynamic weight update mechanism: adjust the weight of federal model according to the data value degree, and the update formula is, In the formula, ω i is the weight coefficient of the i-th environmental parameter; T exposure is the data exposure time; i is the index of the federated learning participant.

5. The emergency rescue data privacy protection and collaborative analysis system based on federated learning according to claim 1, characterized in that, The blockchain smart contract module comprises: Data life cycle audit unit: used for recording the whole process operation from data collection to destruction; Self-destruction key mechanism: automatically destroy the key when illegal access is detected; Permission dynamic adjustment unit: assign data access permissions according to the trust level of the device.

6. The emergency rescue data privacy protection and collaborative analysis system based on federated learning according to claim 1, characterized in that, Further comprising a trusted group construction unit, the trusted group construction unit dynamically generates a task-driven device cluster based on device cooperation trust degree, and the overall trust degree of the cluster is calculated as: In the formula, T group is the overall credibility of the device cluster; M is the total number of devices in the cluster; k is the index number of the device in the cluster; is the dynamic credibility of the kth device; φ k is the weight coefficient of the kth device.

7. The emergency rescue data privacy protection and collaborative analysis system based on federated learning according to claim 1, characterized in that, Further comprising: Lightweight cryptography module: data encryption is realized by using SM4-GCM of national secret algorithm; Key security management unit: session key is generated and stored in trusted execution environment.

8. The emergency rescue data privacy protection and collaborative analysis system based on federated learning according to claim 7, characterized in that, The lightweight cryptography module comprises: SM2 algorithm generates a key pair, and the key length is 4096 bits; SM3 calculates data hash value; SM4-CTR mode encrypts data stream.

9. The emergency rescue data privacy protection and collaborative analysis system based on federated learning according to claim 1, characterized in that, Further comprising an emergency resource optimization scheduling module, the emergency resource optimization scheduling module realizes resource allocation by improving differential evolution algorithm, specifically comprising: Establish resource scheduling model: In the formula, V is the emergency resource scheduling priority score; N is the total number of resource types participating in scheduling; CVSS i is the vulnerability threat score of the ith type of resource; ω i is the weight coefficient of the ith type of resource; λ is the time decay coefficient; T i is the idle time of the ith type of resource; e≈2.718 is a natural constant. Visual decision support: based on GIS map to show resource distribution and scheduling scheme.

10. The federated learning based emergency rescue data privacy protection and collaborative analysis system according to claim 1, characterized in that, The terminal device comprises a trusted box hardware, and the trusted box hardware comprises: Raspberry Pi master control platform; National secret algorithm cryptography chip; Anti-side channel protection circuit: resist differential power analysis by dynamic noise injection; Trusted boot mechanism: verify Bootloader, kernel and application layer hash value step by step.

Images