An AI-driven information management method
By combining deep adaptive probabilistic clustering and ECDH-SHA-512 hybrid encryption technology with blockchain and Ethereum smart contracts, we have achieved efficient, secure and reliable data management for enterprise IoT fog node clusters. This solves the problems of unstable fog node cluster management and data transmission security, and provides fine-grained control over cross-domain access.
Patent Information
- Application Number
- CN202611227365.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-13
- Publication Date
- 2026-11-17
AI Technical Summary
Fog node clusters in enterprise IoT lack adaptive optimization mechanisms, and cluster head selection is arbitrary, resulting in unstable network topology, low data transmission efficiency, lack of lightweight encryption and integrity verification for sensitive data transmission, vulnerability to man-in-the-middle attacks, single point of failure risk in identity authentication and access control, and rudimentary cross-departmental and cross-regional access control, making log traceability difficult.
A deep adaptive probabilistic clustering algorithm is used to optimize the fog node cluster and select the optimal cluster head. Lightweight key negotiation and data integrity protection are achieved through ECDH-SHA-512 hybrid encryption. Authentication records are stored in a decentralized manner based on blockchain. Ethereum smart contracts are used to automatically execute authentication and authorization. Trusted cross-domain access control is achieved by combining a master-slave chain architecture. Intrusion detection is performed in real time at the edge of the fog node.
It achieves efficient and stable management of fog node clusters, ensures the security and integrity of end-to-end encrypted transmission of sensitive enterprise data, provides tamper-proof data storage, and enables fine-grained control of cross-domain access permissions, thus solving the security and efficiency pain points in traditional enterprise information management.
Smart Images

Figure CN123122886A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of enterprise information management technology, specifically to an artificial intelligence-driven information management method. Background Technology
[0002] As enterprises deepen their digital transformation, IoT terminals and edge computing nodes have become core carriers for enterprise data collection, transmission, and management. The security protection of sensitive and private data such as employee identity information, customer data, supply chain data, and production operation parameters is directly related to the protection of enterprise trade secrets and compliance governance. However, in existing technologies, fog node clusters in enterprise IoT lack adaptive optimization mechanisms, and arbitrary cluster head selection leads to unstable network topology and low data transmission efficiency. Sensitive enterprise data lacks lightweight, end-to-end encryption and integrity verification methods during transmission, making it vulnerable to man-in-the-middle attacks or data tampering. Identity authentication and access control often adopt centralized architectures, posing single-point-of-failure risks. Access control across departments, regions, and third-party partners is often lax, and log traceability is difficult. Summary of the Invention
[0003] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides an AI-driven information management method. Addressing the technical problems of disordered management of fog node clusters at the IoT terminal edge, lack of adaptive optimization in cluster head selection, and vulnerability to the theft and tampering of sensitive enterprise data transmission, this solution employs a deep adaptive probabilistic clustering algorithm to optimize the fog node cluster, selects the optimal cluster head, and uses ECDH-SHA-512 hybrid encryption to achieve lightweight key negotiation and data integrity protection. Addressing the technical problems of lax cross-domain identity authentication and access control, and the difficulty in managing third-party access permissions and tracing audits, this solution is based on decentralized storage of authentication records on blockchain, with Ethereum smart contracts automatically executing authentication and authorization. A master-slave chain architecture is used to achieve trusted cross-domain access control, and intrusion detection is performed in real time at the fog node edge. Abnormal information is stored on the blockchain as evidence, balancing edge efficiency, lightweight security, and decentralized trust, thus solving the pain points of security and efficiency in enterprise information storage, authentication, and access.
[0004] The technical solution adopted by this invention is as follows: This invention provides an artificial intelligence-driven information management method, which specifically includes the following steps:
[0005] Step S1: Enterprise data upload. Collect enterprise data and encapsulate it into standard data frames via the MQTT IoT protocol, then transmit it to the input port of the fog node cluster management. The enterprise data includes employee personal information, customer information, and terminal device operation data. Employee personal information includes name, employee number, contact information, department, attendance records, and performance data. Terminal device information includes device number, model, deployment location, and operating status.
[0006] Step S2: Fog node cluster management. Using enterprise edge gateways and IoT terminals as fog nodes, a deep adaptive probabilistic clustering model is used to dynamically manage the fog nodes and dynamically select the fog node cluster head. The standardized and noise-reduced terminal-collected data is then mixed and encrypted, while providing an edge deployment environment for intrusion detection.
[0007] Step S3: Hybrid encryption, combining ECDH elliptic curve key exchange and SHA-512 secure hash algorithm to perform end-to-end encryption and integrity verification on the terminal-collected data to obtain encrypted secure data;
[0008] Step S4: Blockchain authentication, which records encrypted and secure data on the blockchain and stores authentication records and access logs;
[0009] Step S5: Smart contract, based on Ethereum contract to solidify security rules, automatically execute authentication, access authorization and intrusion response, and generate authorization tokens for cross-domain access control;
[0010] Step S6: Cross-domain access control, using a master-slave chain structure, uses authorization tokens to control access between fog nodes and third parties;
[0011] Step S7: Intrusion detection. Relying on the fog node cluster, real-time edge detection of anomalies is performed. Once an anomaly is detected, the smart contract is immediately notified to trigger a response. The alarm information is synchronously transmitted to the blockchain for on-chain evidence storage, forming a closed loop of detection, response, and evidence storage.
[0012] Further, step S2, fog node cluster management, specifically includes:
[0013] Step S21: Deep adaptive clustering provides dynamic threshold and objective function optimization to achieve intelligent extraction of fog node features, reduce clustering noise, and improve cluster stability. The specific operation is as follows:
[0014] A dynamic threshold mechanism is set up. Dynamic upper and lower thresholds are generated through clustering DBI index iterative optimization. Fog node feature vectors are extracted, and the cosine similarity between every two fog node feature pairs is calculated to construct a similarity matrix. Data with similarity values within the threshold range are marked as invalid data and removed to avoid interference from noisy data.
[0015] Objective function optimization: Construct the objective function of the deep adaptive probabilistic clustering model. The objective function is the difference between the dynamic upper threshold and the dynamic lower threshold. The objective function is optimized using the gradient descent algorithm to reduce the clustering error round by round. The sample selection range is controlled by adaptive parameters. The adaptive parameters are iteratively updated using the learning rate. During training, the adaptive parameters are gradually increased to optimize the cluster heads.
[0016] A hybrid training strategy is adopted, which generates 0 / 1 pseudo-labels for fog node pairs based on the similarity matrix and dynamic upper and lower thresholds, calculates the self-labeling matrix, calculates the similarity loss based on the similarity matrix and the self-labeling matrix, and iteratively trains the deep adaptive probabilistic clustering model. The iteration stops when the dynamic upper threshold is less than or equal to the dynamic lower threshold.
[0017] Step S22: Calculate the optimal probability of probabilistic clustering. Based on the characteristics of fog nodes, calculate the optimal selection probability of cluster head. Specifically, fog nodes follow a uniform spatial Poisson process. Based on the energy consumption, transmission bandwidth and communication distance of fog nodes, substitute them into the Poisson probability model to calculate the probability of fog nodes becoming cluster heads.
[0018] Calculate the average distance from fog nodes to the cluster head processing center. Preset the initial probability of fog nodes becoming cluster heads. Then, the cluster head node strength is the product of the original node strength and the initial probability, and the non-cluster head node strength is the product of the original node strength and the non-cluster head initial probability. The original node strength is the fog node density. Connect the non-cluster head nodes to the nearest cluster head to form a Voronoi cell. Calculate the node expectation and total link length expectation within the Voronoi cell to obtain the optimal cluster head selection probability.
[0019] Step S23: Optimal cluster head selection. A weighted fusion mechanism is adopted to sum the similarity weight of deep adaptive clustering with the cluster head probability weight calculated by probability calculation, and select the fog node with the highest weight as the dynamic cluster head.
[0020] Further, step S3, hybrid encryption, specifically includes:
[0021] Key exchange is achieved through elliptic curve Diffie-Hellman, enabling key negotiation and data encryption.
[0022] Hash verification uses the SHA-512 algorithm to generate a checksum, ensuring that the data is not tampered with.
[0023] The key exchange specifically involves setting elliptic curve basic parameters, introducing a certificate authority to issue trusted digital signatures for each IoT terminal, and when IoT terminals communicate with each other, each party generates an ECDH key pair. After exchanging public keys, both parties independently calculate a shared session key, completing secure key negotiation without transmitting a key. Enterprise data is then transmitted in encrypted form. After decryption, a new shared key is negotiated with the edge gateway, and the data is forwarded in encrypted form. The edge gateway authenticates the forwarded data using a CA digital signature and then uploads it to the enterprise data center, achieving end-to-end encryption.
[0024] The hash verification specifically involves padding the message length and appending the original message length value, setting an initial hexadecimal hash value as the basis for iterative calculation, and generating a message digest for hash verification.
[0025] Further, in step S4, blockchain authentication, based on the decentralized ledger mechanism of blockchain fog nodes, all historical authentication records, intrusion events, alarm information, and security response operations are written into the blockchain ledger. Fog nodes collectively verify intrusion behavior and authentication results through the PoW / PoS consensus mechanism to resist malicious nodes from tampering with data. Fog nodes are deployed at the network edge to directly collect data from enterprise edge gateway devices and detect intrusion behavior in real time.
[0026] Further, in step S5, the smart contract is based on Ethereum smart contract automation. The Ethereum smart contract is written in the Solidity Turing-complete language, compiled into Ethereum Virtual Machine bytecode, and then uploaded to the blockchain to obtain a unique contract address. The identity authentication rules are automatically executed. If the verification is successful, access is authorized; if it fails, access is denied. The access control policy is automatically executed to finely manage third-party and cross-domain access permissions, and to respond to intrusion attempts.
[0027] Further, step S6, cross-domain access control, adopts a main chain and domain slave chain architecture. The main chain is responsible for cross-domain trust verification, and the slave chain is responsible for device management within a single domain, realizing the decentralization and efficiency of cross-domain authentication. Specifically, it includes the following steps:
[0028] Step S61: Cross-domain request initiation. When a fog node migrates from the source domain to the target domain, the fog node initiates a cross-chain authentication request to the target domain's sub-chain.
[0029] Step S62: The target domain sends a random number challenge from the chain. The target domain generates a random number from the chain and sends it to the global main chain and the fog node respectively. The domain credential of the fog node is hashed.
[0030] Step S63: Result comparison and verification. The global main chain and fog node return the calculation results. The target domain's sub-chain compares the results for consistency. If the results are consistent, the fog node is determined to be trustworthy, the target domain system access interface is opened, and the authentication record is stored on the chain. If the results are inconsistent, the fog node is determined to be untrustworthy, and the access request is rejected.
[0031] Step S64: Access audit traceability. All cross-domain authentication and access records are permanently stored on the blockchain to support post-event auditing and accountability.
[0032] The beneficial effects achieved by the present invention using the above solution are as follows:
[0033] (1) The edge cluster is efficient and stable. It realizes dynamic cluster management of fog nodes through deep adaptive probabilistic clustering model. Combined with dynamic threshold screening, objective function optimization and hybrid training strategy, it effectively eliminates clustering noise, intelligently selects cluster head nodes, and improves the standardization and reliability of terminal data collection and transmission.
[0034] (2) Data transmission is secure and reliable. It integrates ECDH elliptic curve key exchange and SHA-512 hash algorithm to realize end-to-end encrypted transmission and integrity verification of sensitive enterprise data. It can complete the security key negotiation without plaintext transmission key, prevent data theft, tampering and forgery from the source, and ensure data security throughout the entire chain.
[0035] (3) Data storage is tamper-proof. Relying on the blockchain decentralized ledger and PoW / PoS consensus mechanism, the authentication records, access logs, intrusion alarms and response operations are all stored on the chain, completely eliminating the risk of malicious nodes tampering with data. The master-slave chain architecture is adopted to realize cross-domain trust verification and single-domain device management. Cross-domain trusted authentication is completed by combining random number challenge and hash comparison. The access permissions of third parties and cross-domain access are finely controlled, solving the problems of crude cross-domain access control and lack of trust in the traditional cross-domain access control. Attached Figure Description
[0036] Figure 1 The present invention provides a flowchart of the steps of an artificial intelligence-driven information management method.
[0037] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation
[0038] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0039] Example 1: See Figure 1This embodiment provides an artificial intelligence-driven information management method, which specifically includes the following steps:
[0040] Step S1: Enterprise data upload. Collect enterprise data and encapsulate it into standard data frames via the MQTT IoT protocol, then transmit it to the input port of the fog node cluster management. The enterprise data includes employee personal information, customer information, and terminal device operation data. Employee personal information includes name, employee number, contact information, department, attendance records, and performance data. Terminal device information includes device number, model, deployment location, and operating status.
[0041] Step S2: Fog node cluster management. Using enterprise edge gateways and IoT terminals as fog nodes, a deep adaptive probabilistic clustering model is used to dynamically manage the fog nodes and dynamically select the fog node cluster head. The standardized and noise-reduced terminal-collected data is then mixed and encrypted, while providing an edge deployment environment for intrusion detection.
[0042] Step S3: Hybrid encryption, combining ECDH elliptic curve key exchange and SHA-512 secure hash algorithm to perform end-to-end encryption and integrity verification on the terminal-collected data to obtain encrypted secure data;
[0043] Step S4: Blockchain authentication, which records encrypted and secure data on the blockchain and stores authentication records and access logs;
[0044] Step S5: Smart contract, based on Ethereum contract to solidify security rules, automatically execute authentication, access authorization and intrusion response, and generate authorization tokens for cross-domain access control;
[0045] Step S6: Cross-domain access control, using a master-slave chain structure, uses authorization tokens to control access between fog nodes and third parties;
[0046] Step S7: Intrusion detection. Relying on the fog node cluster, real-time edge detection of anomalies is performed. Once an anomaly is detected, the smart contract is immediately notified to trigger a response. The alarm information is synchronously transmitted to the blockchain for on-chain evidence storage, forming a closed loop of detection, response, and evidence storage.
[0047] Example 2: See Figure 1 This embodiment is based on the above embodiment. Step S2, fog node cluster management, specifically includes:
[0048] Step S21: Deep adaptive clustering provides dynamic threshold and objective function optimization to achieve intelligent extraction of fog node features, reduce clustering noise, and improve cluster stability. The specific operation is as follows:
[0049] A dynamic threshold mechanism was set up, and based on the measured noise distribution of network edge nodes and the iterative optimization of clustering effectiveness index, the simulation verified that the upper dynamic threshold of 0.85 and the lower dynamic threshold of 0.3 can remove noise to the greatest extent and retain effective clustering data. Fog node feature vectors were extracted, and the cosine similarity between every two fog node feature pairs was calculated to construct a similarity matrix. Data with similarity values within the threshold range were marked as invalid data and removed to avoid interference from noise data.
[0050] Objective function optimization: The objective function of the deep adaptive probabilistic clustering model is constructed using gradient descent to optimize the objective function, progressively reducing the clustering error. The sample selection range is controlled by adaptive parameters, which are iteratively updated using a learning rate. During training, the adaptive parameters are gradually increased to optimize the cluster heads. The formula used is as follows: ;
[0051] In the formula, It is an adaptive parameter. This is the current iteration number, initially set to 0.5. This is the learning rate, initially set to 0.01. It is the objective function to be minimized, which is the difference between the dynamic upper threshold and the dynamic lower threshold;
[0052] A hybrid training strategy is used to generate 0 / 1 pseudo-labels for fog node pairs based on the similarity matrix and dynamic upper and lower thresholds. Similarity values greater than the dynamic upper threshold are marked as 1, and similarity values less than the dynamic lower threshold are marked as 0, resulting in a self-labeling matrix. The similarity loss is calculated based on the similarity matrix and the self-labeling matrix, and the deep adaptive probabilistic clustering model is iteratively trained. The iteration stops when the dynamic upper threshold is less than or equal to the dynamic lower threshold.
[0053] Step S22: Calculate the optimal probability of probabilistic clustering. Based on the characteristics of fog nodes, calculate the optimal selection probability of cluster head. Specifically, fog nodes follow a uniform spatial Poisson process. Based on the energy consumption, transmission bandwidth and communication distance of fog nodes, substitute them into the Poisson probability model to calculate the probability of fog nodes becoming cluster heads.
[0054] Calculate the average distance from fog nodes to the cluster head processing center. Preset the initial probability of fog nodes becoming cluster heads. Then, the cluster head node strength is the product of the original node strength and the initial probability, and the non-cluster head node strength is the product of the original node strength and the non-cluster head initial probability. The original node strength is the fog node density. Connect the non-cluster head nodes to the nearest cluster head to form a Voronoi cell. Calculate the node expectation and total link length expectation within the Voronoi cell to obtain the optimal cluster head selection probability.
[0055] Step S23: Optimal cluster head selection. A weighted fusion mechanism is adopted to sum the similarity weight of deep adaptive clustering with the cluster head probability weight calculated by probability calculation, and select the fog node with the highest weight as the dynamic cluster head.
[0056] Example 3: See Figure 1 This embodiment is based on the above embodiment. Step S3, hybrid encryption, specifically includes:
[0057] Key exchange is achieved through elliptic curve Diffie-Hellman, enabling key negotiation and data encryption.
[0058] Hash verification uses the SHA-512 algorithm to generate a checksum, ensuring that the data is not tampered with.
[0059] The key exchange specifically involves setting elliptic curve basic parameters, introducing a certificate authority to issue trusted digital signatures for each IoT terminal, and when IoT terminals communicate with each other, each party generates an ECDH key pair. After exchanging public keys, both parties independently calculate a shared session key, completing secure key negotiation without transmitting a key. Enterprise data is then transmitted in encrypted form. After decryption, a new shared key is negotiated with the edge gateway, and the data is forwarded in encrypted form. The edge gateway authenticates the forwarded data using a CA digital signature and then uploads it to the enterprise data center, achieving end-to-end encryption.
[0060] The hash verification process involves padding the message length to 896 mod 1024, appending 128 bits of the original message length value, and if the message is too long, taking mod 2 to the power of 128. Eight 64-bit hexadecimal initial hash values are set as the basis for iterative calculation. Data is processed iteratively in 1024-bit blocks to finally generate a message digest for hash verification.
[0061] Example 4: See Figure 1 This embodiment is based on the above embodiment. In step S4, blockchain authentication is performed using the decentralized ledger mechanism of blockchain fog nodes. All historical authentication records, intrusion events, alarm information, and security response operations are written into the blockchain ledger. Fog nodes use PoW / PoS consensus mechanisms. In PoW mode, the ledger node is elected through computing power competition, and in PoS mode, the ledger node is elected through node staking weight. After the signature is verified across the entire network, it is uploaded to the chain to resist malicious tampering. Fog nodes are deployed on the edge network and directly connected to the blood collection terminal via RS485 / Bluetooth to collect identity authentication, access requests, and device status data in real time. Intrusion detection is completed based on the security rule base and traffic thresholds. Security rules are matched, and if abnormal information is found, it is immediately uploaded to the chain for evidence storage.
[0062] Example 5: See Figure 1This embodiment is based on the above embodiment. In step S5, the smart contract is based on Ethereum smart contract automation. The Ethereum smart contract is written in the Solidity Turing complete language, compiled into Ethereum Virtual Machine bytecode and then put on the chain to obtain a unique contract address. The identity authentication rules are automatically executed. If the verification is successful, access is authorized; if it fails, access is denied. The access control policy is automatically executed to finely manage third-party and cross-domain access permissions, and at the same time respond to intrusion behavior.
[0063] Example 6: See Figure 1 This embodiment is based on the above embodiment. Step S6, cross-domain access control, adopts a main chain and domain slave chain architecture. The main chain is responsible for cross-domain trust verification, and the slave chain is responsible for device management within a single domain, realizing the decentralization and efficiency of cross-domain authentication. Specifically, it includes the following steps:
[0064] Step S61: Cross-domain request initiation. When a fog node migrates from the source domain to the target domain, the fog node initiates a cross-chain authentication request to the target domain's sub-chain. The request text format is as follows: ;
[0065] In the formula, It is a cross-domain authentication request. It is the public key of the fog node's device. It is the private key signature of the fog node. It is the public key of the source domain chain. It is the domain credential of the fog node. It is the target access object. It is a hash calculation request. This includes additional device information for fog nodes, such as model number and service permissions. It's a timestamp;
[0066] Step S62: The target domain sends a random number challenge from the chain. The target domain generates a random number from the chain and sends it to the global main chain and the fog node respectively. The domain credential of the fog node is hashed.
[0067] Step S63: Result comparison and verification. The global main chain and fog node return the calculation results. The target domain's sub-chain compares the results for consistency. If the results are consistent, the fog node is determined to be trustworthy, the target domain system access interface is opened, and the authentication record is stored on the chain. If the results are inconsistent, the fog node is determined to be untrustworthy, and the access request is rejected.
[0068] Step S64: Access audit traceability. All cross-domain authentication and access records are permanently stored on the blockchain to support post-event auditing and accountability.
[0069] Example 7: See Figure 1This embodiment is based on the above embodiment. Cross-domain access control uses the Transport Layer Security (TLS) protocol to ensure cross-domain communication security. Data is encrypted throughout the storage and transmission process, providing a guarantee for secure transmission.
[0070] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. It is understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the invention. The scope of the invention is defined by the appended claims and their equivalents. If those skilled in the art, inspired by this description, design similar structural methods and embodiments without departing from the inventive spirit of the invention, such designs should fall within the protection scope of the invention.
Claims
1. An artificial intelligence-driven information management method, characterized in that, Specifically, the following steps are included: Step S1: Enterprise data upload. Collect enterprise data, encapsulate it into standard data frames using the MQTT IoT protocol, and transmit it to the input port of the fog node cluster management. Step S2: Fog node cluster management. Using enterprise edge gateways and IoT terminals as fog nodes, a deep adaptive probabilistic clustering model is used to dynamically manage the fog nodes and dynamically select the fog node cluster head to obtain the terminal-collected data. Step S3: Hybrid encryption, combining ECDH elliptic curve key exchange and SHA-512 secure hash algorithm to perform end-to-end encryption and integrity verification on the terminal-collected data to obtain encrypted secure data; Step S4: Blockchain authentication, which records encrypted and secure data on the blockchain and stores authentication records and access logs; Step S5: Smart contract, based on Ethereum contract to solidify security rules, automatically execute authentication, access authorization and intrusion response, and generate authorization tokens for cross-domain access control; Step S6: Cross-domain access control, using a master-slave chain structure, uses authorization tokens to control access between fog nodes and third parties; Step S7: Intrusion detection. Relying on the fog node cluster, real-time edge detection of anomalies is performed. Once an anomaly is detected, the smart contract is immediately notified to trigger a response. The alarm information is synchronously transmitted to the blockchain for on-chain evidence storage, forming a closed loop of detection, response, and evidence storage.
2. The artificial intelligence-driven information management method according to claim 1, characterized in that, Step S2, Fog Node Cluster Management, specifically includes: Step S21: Deep adaptive clustering provides dynamic threshold and objective function optimization to intelligently extract fog node features; Step S22: Calculate the optimal probability of probabilistic clustering. Based on the characteristics of fog nodes, calculate the optimal selection probability of the cluster head. Step S23: Optimal cluster head selection. A weighted fusion mechanism is adopted to sum the similarity weight of deep adaptive clustering with the cluster head probability weight calculated by probability calculation, and select the fog node with the highest weight as the dynamic cluster head.
3. The artificial intelligence-driven information management method according to claim 2, characterized in that, Step S21, deep adaptive clustering, the specific operation is as follows: A dynamic threshold mechanism is set up. Dynamic upper and lower thresholds are generated through clustering DBI index iterative optimization. Fog node feature vectors are extracted, and the cosine similarity between every two fog node feature pairs is calculated to construct a similarity matrix. Data with similarity values within the threshold range are marked as invalid data and removed to avoid interference from noisy data. Objective function optimization: Construct the objective function of the deep adaptive probabilistic clustering model. The objective function is the difference between the dynamic upper threshold and the dynamic lower threshold. The objective function is optimized using the gradient descent algorithm to reduce the clustering error round by round. The sample selection range is controlled by adaptive parameters. The adaptive parameters are iteratively updated using the learning rate. During training, the adaptive parameters are gradually increased to optimize the cluster heads. A hybrid training strategy is adopted, which generates 0 / 1 pseudo-labels for fog node pairs based on the similarity matrix and dynamic upper and lower thresholds, calculates the self-labeling matrix, calculates the similarity loss based on the similarity matrix and the self-labeling matrix, and iteratively trains the deep adaptive probabilistic clustering model. The iteration stops when the dynamic upper threshold is less than or equal to the dynamic lower threshold.
4. The artificial intelligence-driven information management method according to claim 3, characterized in that, Objective function optimization: The objective function of the deep adaptive probabilistic clustering model is constructed, and the adaptive parameters are updated iteratively using the learning rate. The formula used is as follows: ; In the formula, It is an adaptive parameter. This is the current iteration number, initially set to 0.
5. This is the learning rate, initially set to 0.
01. It is the objective function to be minimized, which is the difference between the dynamic upper threshold and the dynamic lower threshold.
5. The artificial intelligence-driven information management method according to claim 3, characterized in that, Step S22, Calculate the optimal probability of probabilistic clustering. Specifically, fog nodes follow a uniform spatial Poisson process. Based on the fog node's energy consumption, transmission bandwidth, and communication distance, substitute them into the Poisson probability model to calculate the probability that the fog node becomes the cluster head. Calculate the average distance from fog nodes to the cluster head processing center. Preset the initial probability of fog nodes becoming cluster heads. Then, the cluster head node strength is the product of the original node strength and the initial probability, and the non-cluster head node strength is the product of the original node strength and the non-cluster head initial probability. The original node strength is the fog node density. Connect the non-cluster head nodes to the nearest cluster head to form a Voronoi cell. Calculate the expected node length and total link length within the Voronoi cell to obtain the optimal cluster head selection probability.
6. The artificial intelligence-driven information management method according to claim 1, characterized in that, Step S3, mixed encryption, specifically includes: Key exchange is achieved through elliptic curve Diffie-Hellman, enabling key negotiation and data encryption. Hash verification uses the SHA-512 algorithm to generate a checksum, ensuring that the data is not tampered with.
7. The artificial intelligence-driven information management method according to claim 6, characterized in that, The key exchange specifically involves setting elliptic curve basic parameters, introducing a certificate authority to issue trusted digital signatures for each IoT terminal, and when IoT terminals communicate with each other, each party generates an ECDH key pair. After exchanging public keys, both parties independently calculate a shared session key, completing secure key negotiation without transmitting a key. Enterprise data is then transmitted in encrypted form. After decryption, a new shared key is negotiated with the edge gateway, and the data is forwarded in encrypted form. The edge gateway authenticates the forwarded data using a CA digital signature and then uploads it to the enterprise data center, achieving end-to-end encryption. The hash verification specifically involves padding the message length and appending the original message length value, setting an initial hexadecimal hash value as the basis for iterative calculation, and generating a message digest for hash verification.
8. The artificial intelligence-driven information management method according to claim 1, characterized in that, Step S4, Blockchain Authentication: Based on the decentralized ledger mechanism of blockchain fog nodes, all historical authentication records, intrusion events, alarm information, and security response operations are written into the blockchain ledger. Fog nodes collectively verify intrusion behavior and authentication results through the PoW / PoS consensus mechanism. Fog nodes are deployed at the network edge to directly collect data from enterprise edge gateway devices and detect intrusion behavior in real time.
9. The artificial intelligence-driven information management method according to claim 1, characterized in that, Step S5, Smart Contract: Based on Ethereum smart contract automation, the Ethereum smart contract is written in the Solidity Turing-complete language, compiled into Ethereum Virtual Machine bytecode and then uploaded to the blockchain to obtain a unique contract address. The identity authentication rules are automatically executed. If the verification is successful, access is authorized; if it fails, access is denied. The access control policy is automatically executed to finely manage third-party and cross-domain access permissions, and to respond to intrusion attempts.
10. The artificial intelligence-driven information management method according to claim 1, characterized in that, Step S6, cross-domain access control, adopts a main chain and domain slave chain architecture. The main chain is responsible for cross-domain trust verification, and the slave chain is responsible for device management within a single domain, realizing the decentralization and efficiency of cross-domain authentication. Specifically, it includes the following steps: Step S61: Cross-domain request initiation. When a fog node migrates from the source domain to the target domain, the fog node initiates a cross-chain authentication request to the target domain's sub-chain. Step S62: The target domain sends a random number challenge from the chain. The target domain generates a random number from the chain and sends it to the global main chain and the fog node respectively. The domain credential of the fog node is hashed. Step S63: Result comparison and verification. The global main chain and fog node return the calculation results. The target domain's sub-chain compares the results for consistency. If the results are consistent, the fog node is determined to be trustworthy, the target domain system access interface is opened, and the authentication record is stored on the chain. If the results are inconsistent, the fog node is determined to be untrustworthy, and the access request is rejected. Step S64: Access audit traceability. All cross-domain authentication and access records are permanently stored on the blockchain to support post-event auditing and accountability.
