Internet of Things security data management method, medium and system based on artificial intelligence

Abstract

The invention belongs to the field of artificial intelligence and the technical field of Internet of Things security, and discloses an Internet of Things security data management method based on artificial intelligence, a medium and a system, and the related entities comprise six types of entities: a data owner (DO), a data visitor (DU), an attribute authorization center (AS), a block chain (BC), Internet of Things equipment (IOT) and an IPFS network (IPFS). The scheme comprises six processes. According to the Internet of Things security data management method based on artificial intelligence, the medium and the system disclosed by the invention, the block chain technology is fused, and the Fabric alliance chain, the PBFT consensus algorithm and the random forest evaluation model are introduced, so that the auditing performance in the Internet of Things data transmission and cloud storage processes is realized, the data integrity and decryption correctness are ensured, and the security of the Internet of Things is improved. And the opaqueness problem caused by the untrusted third-party Internet of Things terminal is effectively solved. In addition, an IPFS system is further introduced, storage of IOT data is achieved, and the storage pressure of the block chain is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of artificial intelligence and the field of Internet of Things security technology, and is oriented to the background of Internet of Things data cloud storage. BACKGROUND

[0002] With the rapid development of Internet of Things (IoT) technology, intelligent devices have been widely penetrated into key fields such as smart cities, industrial control, smart homes and medical health, but the subsequent infringement attacks on IoT data have also increased day by day. Malicious traffic attacks (such as DDoS attacks, Sybil attacks, data theft, malware intrusion and botnet spread) targeting IoT are showing the evolution trend of scale, concealment and intelligence. Therefore, protecting the security of IoT data has become the main object of research. The application realizes the safe cloud management of IoT data by means of blockchain technology and artificial intelligence technology.

[0003] Blockchain is a decentralized, tamper-proof, traceable, multi-party distributed database blockchain that integrates P2P protocol, asymmetric encryption technology, consensus mechanism and other technologies. As long as the honest nodes have the advantage of CPU computing power, the system is safe and reliable. Therefore, the security mechanism of blockchain can effectively solve the data authenticity problem in IoT applications and to some extent guarantee the confidentiality and availability of IoT data. In addition, due to the exponential growth of IOT related data, the storage space required by the blockchain system will become larger and larger. If all the data generated by the cloud device is stored on the blockchain, the size of the entire blockchain will be very large, so we only store the data access key on the blockchain and store the encrypted data in the cloud file system. Reduce the burden of the blockchain system.

[0004] At the same time, by using an artificial intelligence model, the IoT traffic data is preprocessed to enhance and prevent malicious traffic from being missed due to a small proportion; the Gini impurity reduction of random forest is used as a feature importance measure; to solve the limitations of static feature weights, a multi-head attention mechanism is introduced to dynamically adjust feature importance; federated learning is used for distributed training; and a lightweight model is deployed. SUMMARY

[0005] The technical scheme adopted by the scheme of the application is: an IoT security data management method, medium and system based on artificial intelligence. The entities involved include six types of entities: data owner (DO), data user (DU), attribute authorization center (AS), blockchain (BC), IoT device (IOT), and IPFS network (IPFS).

[0006] An IoT security data management method, medium and system based on artificial intelligence, characterized by comprising the following steps:

[0007] Step 1: System initial configuration. Complete key initialization, model training;

[0008] Step 2: Block generation. Generate main blocks and obtain auxiliary ciphertext;

[0009] Step 3: Data detection and encryption. Detect malicious traffic using an artificial intelligence model and encrypt non-malicious traffic data;

[0010] Step 4: Dynamic adjustment. Dynamically adjust model thresholds;

[0011] Step 5: Download ciphertext. Interact with BC and IPFS to obtain ciphertext;

[0012] Step 6: Decryption. DU decrypts to obtain plaintext;

[0013] Further, step 1 includes the following specific steps:

[0014] Step 1.1: Each client of the system registers and obtains an identity certificate by calling the certificate CA service through the SDK. Therefore, DU, DO, and AA also have their public and private key pairs in Fabric;

[0015] DO calculates and broadcasts it to the BC node;

[0016] Step 1.2: AS obtains sufficient ;

[0017] Step 1.3: AS generates attribute base main public key MPK and main private key MSK upon receiving a task, and generates decryption key DK for potential DU, sends MPK to DO, and exchanges key KDO with DO, waiting for DK request;

[0018] Step 1.4: AS uses the Synthetic Minority Oversampling Technique (SMOTE) algorithm for data enhancement processing, synthesizing sample interpolation . For a random sample in the minority class sample set, k-nearest neighbor samples (k=5), a random number of 0 or 1 controlling the position of the synthesized sample, a new minority class sample generated in the feature space;

[0019] Step 1.5: Introduce an adaptive near-neighbor selection mechanism, use k=3 near-neighbor for edge samples (Near-miss samples) and k=7 for safe samples, to solve the over-generalization problem of traditional SMOTE;

[0020] Step 1.6: Use the reduction in Gini impurity of random forest as a measure of feature importance: The feature importance scoring algorithm is as follows: Where n represents the number of node samples and K represents the number of decision trees. This indicates whether a node is split. Through calculation, the original 42-dimensional features are reduced to 6-dimensional core features;

[0021] Step 1.7: To address the limitations of static feature weights, a multi-head attention mechanism is introduced to dynamically adjust feature importance:

[0022] Step 1.8: Use As a bootstrap sampling algorithm, feature random selection ;

[0023] Step 1.9: To address the data privacy issue in the Internet of Things (IoT), a federated averaging algorithm is used to implement distributed training. , This is the global model after t iterations. For learning rate, Update the local model;

[0024] Furthermore, step 2 includes the following specific steps:

[0025] Step 2.1: DO sets up KDO for IOT;

[0026] Step 2.2: Encrypt the key KDO using attribute base encryption to obtain the ciphertext CT;

[0027] Step 2.3: Generate the ciphertext CT for block storage;

[0028] Furthermore, step 3 includes the following specific steps:

[0029] Step 3.1: The IOT obtains the data generated for the DO;

[0030] Step 3.2: Detect whether there is malicious traffic in the data.

[0031] Step 3.3: The IoT uses KDO to encrypt secure data, generating ciphertext CT′, logging malicious traffic, and reporting it.

[0032] Step 3.4: Store CT′ in IPFS;

[0033] Furthermore, step 4 includes the following specific steps:

[0034] Step 4.1: IPFS is responsible for recording storage information and storing it in the BC node;

[0035] Step 4.2: read the log, obtain traffic detection information, and design an adaptive threshold based on the 3σ principle , is an abnormality judgment threshold at time t, is the traffic mean value in a time unit, is a traffic fluctuation degree measure;

[0036] Further, step 5 includes the following specific steps:

[0037] Step 5.1: DU requests ciphertext CT from BC;

[0038] Step 5.2: DU decrypts using decryption key DK to obtain KDO;

[0039] Further, step 6 includes the following specific steps:

[0040] Step 6.1: DU downloads ciphertext CT' from IPFS, decrypts using KDO, and obtains data plaintext that meets the access policy;

[0041] Some advantages of the present application are as follows:

[0042] First, short key, reduce consumption;

[0043] Second, in access control, fine-grained access can improve access efficiency;

[0044] Third, IPFS distributed storage data, reduce the storage pressure of BC.

[0045] Fourth, use dynamic threshold adjustment to detect malicious traffic, and the algorithm design is more reasonable. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 is a system flowchart of the present application DETAILED DESCRIPTION

[0047] In order to better understand the purpose, structure and function of the present application, the present application a kind of based on artificial intelligence's Internet of Things security data management method, medium and system will be further described in detail below combining with drawing and implementation example, it should be understood that the implementation example described herein is only for illustration and explanation of the present application, and is not used to limit the present application.

[0048] Example 1

[0049] See Figure 1 The present application provides a kind of based on artificial intelligence's Internet of Things security data management method, medium and system, the entity involved in system includes any data owner, any data visitor, a group of cloud servers (CS), a management center (AA).

[0050] An example of an AI-based IoT secure data management method, medium, and system includes the following steps:

[0051] Step 1: Initial System Configuration. This involves initializing various keys and training the model.

[0052] The following is a further detailed description of step 1 of the present invention:

[0053] Step 1.1: Each client using the system registers and obtains an identity certificate by calling the Certificate CA service through the SDK. Therefore, DU, DO, and AA also have their public and private key pairs in Fabric;

[0054] DO calculates using the system public key And broadcast to the BC node;

[0055] Step 1.2: AS obtains sufficient ;

[0056] Step 1.3: Upon receiving the task, AS generates the attribute base public key MPK and master private key MSK, generates the decryption key DK for the potential DU, sends MPK to DO, exchanges key KDO with DO, and waits for a request for DK.

[0057] Step 1.4: AS uses the Synthetic Minority Oversampling Technique (SMOTE) algorithm for data augmentation, and performs synthetic sample interpolation. . A random sample from the minority class sample set. k nearest neighbor samples (k=5). A random number of 0 or 1 is used to control the position of the synthesized sample. For generating new minority class samples in the feature space;

[0058] Step 1.5: Introduce an adaptive nearest neighbor selection mechanism, using k=3 nearest neighbors for edge samples (near-miss samples) and k=7 for safe samples, to solve the overgeneralization problem of traditional SMOTE;

[0059] Step 1.6: Use the reduction in Gini impurity of random forest as a measure of feature importance: The feature importance scoring algorithm is as follows: Where n represents the number of node samples and K represents the number of decision trees. This indicates whether a node is split. Through calculation, the original 42-dimensional features are reduced to 6-dimensional core features;

[0060] Step 1.7: To solve the limitation of static feature weight, a multi-head attention mechanism is introduced to dynamically adjust feature importance:

[0061] Step 1.8: Use As a self-sampling algorithm, feature random selection ;

[0062] Step 1.9: To solve the problem of data privacy in Internet of Things, a federated average algorithm is used to realize distributed training: , is the global model after t rounds of iteration, is the learning rate, is the local model update;

[0063] Step 2: Block generation. Generate main blocks and obtain auxiliary ciphertext;

[0064] Further detailed description of step 2 of the present application is as follows:

[0065] Step 2.1: DO sets KDO for IOT;

[0066] Step 2.2: Attribute-based encryption is performed on the key KDO to obtain ciphertext CT;

[0067] Step 2.3: Generate block storage ciphertext CT;

[0068] Step 3: Data detection and encryption. Use artificial intelligence model to detect malicious traffic and encrypt non-malicious traffic data;

[0069] Further detailed description of step 3 of the present application is as follows:

[0070] Step 3.1: IOT obtains data generated for DO;

[0071] Step 3.2: Detect whether the data contains malicious traffic.

[0072] Step 3.3: IOT encrypts the secure data using KDO to generate ciphertext CT', logs the malicious traffic, and reports;

[0073] Step 3.4: Store CT' into IPFS;

[0074] Step 4: Dynamic adjustment. Dynamically adjust the model threshold;

[0075] Further detailed description of step 4 of the present application is as follows:

[0076] Step 4.1: IPFS is responsible for recording storage information and recording in BC node;

[0077] Step 4.2: Read the log, get traffic detection information, and design an adaptive threshold based on the 3σ principle , is the abnormality judgment threshold at time t, is the traffic mean value in a time unit, is the traffic fluctuation degree measure;

[0078] Step 5: Download the ciphertext. Interact with BC and IPFS to obtain the ciphertext;

[0079] Further detailed description of step 5 of the present application is as follows:

[0080] Step 5.1: DU requests ciphertext CT from BC;

[0081] Step 5.2: DU decrypts with decryption key DK to obtain KDO;

[0082] Step 6: Decryption. DU decrypts to obtain plaintext;

[0083] Further detailed description of step 6 of the present application is as follows:

[0084] Step 6.1: DU downloads ciphertext CT' from IPFS, decrypts using KDO to obtain data plaintext that meets the access policy

[0085] It should be understood that parts not elaborated in the specification are all prior art.

[0086] It should be understood that the present application is described through some embodiments, and those skilled in the art know that various changes or equivalent replacements can be made to these features and embodiments without departing from the spirit and scope of the present application. In addition, under the teaching of the present application, these features and embodiments can be modified to adapt to specific conditions and materials without departing from the spirit and scope of the present application. Therefore, the present application is not limited by the specific embodiments disclosed herein, and all fall within the scope of the claims of the present application.

Claims

1. An artificial intelligence-based Internet of Things security data management method, medium and system, characterized in that, Comprising the following steps: Step 1: System initial configuration. Complete key initialization, model training; Step 2: Block generation. Generate main blocks and obtain auxiliary ciphertext; Step 3: Data detection and encryption. Detect malicious traffic using an artificial intelligence model and encrypt non-malicious traffic data; Step 4: Dynamic adjustment. Dynamically adjust model thresholds; Step 5: Download ciphertext. Interact with BC and IPFS to obtain ciphertext; Step 6: Decryption. DU decrypts to obtain plaintext. 2.The AI-based IoT security data management method, medium, and system of claim 1, wherein, Step 1 includes the following specific steps: Step 1.1: Each client using the system registers and obtains an identity certificate by calling the certificate CA service through the SDK. Therefore, DU, DO, and AA also have their public and private key pairs in Fabric; The DO calculates with the system public key and broadcasts to the BC nodes; Step 1.2: AS obtains sufficient ; Step 1.3: AS generates attribute-based main public key MPK and main private key MSK upon receiving a task, generates decryption key DK for potential DUs, sends MPK to DO, and exchanges key KDO with DO, waiting for a request for DK; Step 1.4: AS uses the Synthetic Minority Oversampling Technique (SMOTE) algorithm for data augmentation, interpolating synthetic samples . for a random sample in the minority class sample set, k-nearest neighbors of the random sample (k = 5), a random number of 0 or 1 that controls the position of the synthetic sample, is the generated new minority class sample in the feature space; Step 1.5: Introduce an adaptive near-neighbor selection mechanism, using k=3 near-neighbor for edge samples (Near-miss samples) and k=7 for secure samples to solve the over-generalization problem of traditional SMOTE; Step 1.6: Gini impurity reduction using random forest as feature importance measure: with the feature importance score algorithm as: where n denotes the node sample size, K denotes the number of decision trees, and indicates whether the node is split. Through calculation, the 42-dimensional original features are reduced to 6-dimensional core features; Step 1.7: To address the limitations of static feature weights, introduce a multi-head attention mechanism to dynamically adjust feature importance: Step 1.8: Use As a self-sampling algorithm, features are randomly selected ; Step 1.9: To address the issue of data privacy for the Internet of Things, a federated averaging algorithm is used to implement distributed training: , is the global model after t rounds of iteration, is the learning rate, is the local model update. 3.The AI-based IoT security data management method, medium, and system of claim 2, wherein, Step 2 includes the following specific steps: Step 2.1: DO sets KDO for IOT; Step 2.2: Attribute-based encryption is performed on key KDO to obtain ciphertext CT; Step 2.3: Generate block to store ciphertext CT. 4.The AI-based IoT security data management method, medium, and system of claim 3, wherein, Step 3 includes the following specific steps: Step 3.1: IOT obtains data generated for DO; Step 3.2: Detect if there is malicious traffic in the data.

5. Step 3.3: IOT encrypts secure data using KDO to generate ciphertext CT', logs malicious traffic, and reports; Step 3.4: Store CT' in IPFS. 6.The AI-based IoT security data management method, medium, and system of claim 4, wherein, Step 4 includes the following specific steps: Step 4.1: IPFS is responsible for recording and storing information and recording in BC nodes; Step 4.2: Read the log, get traffic detection information, design adaptive threshold based on 3σ principle , is the abnormality judgment threshold at time t, is the traffic mean value in a time unit, is the traffic fluctuation degree measure. 7.The AI-based IoT security data management method, medium, and system of claim 5, wherein, Step 5 includes the following specific steps: Step 5.1: DU requests ciphertext CT from BC; Step 5.2: DU decrypts using decryption key DK to obtain KDO. 8.The AI-based IoT security data management method, medium, and system of claim 6, wherein, Step 6 includes the following specific steps: Step 6.1: DU downloads ciphertext CT' from IPFS and decrypts using KDO to obtain data plaintext that meets the access policy.

Images