Unified framework for trusted computing of artificial intelligence, trusted framework for secure computing of edge devices, security control and decentralization methods

Abstract

The present invention relates to an artificial intelligence trusted computing unified framework based on the PKS system, an artificial intelligence edge device secure computing trusted framework based on the above artificial intelligence trusted computing unified framework, an artificial intelligence edge device security control method based on the above artificial intelligence edge device secure computing trusted framework, and a decentralized edge device trusted authentication method. The present invention can make full use of the native security mechanism of the PKS system to create a full-link security mode from upper-layer applications to underlying operating systems, achieve application security, communication security, data security, and all-round protection of the entire system. At the same time, explore the application of various security mechanisms in the AI ​​industry, ensure the safety and reliability of products, effectively prevent the loss of sensitive and private data, and provide a variety of security protection mechanisms for the future application of artificial intelligence in various industries.

Description

Technical Field

[0001] This patent application belongs to the field of artificial intelligence edge device security control technology. More specifically, it relates to an artificial intelligence trusted computing unified framework based on the PKS system, an edge device security computing trusted framework based on the artificial intelligence trusted computing unified framework, an artificial intelligence edge device security control method, and a decentralized edge device trusted authentication method. Background Art

[0002] Edge devices are devices that provide entry points to the core network of an enterprise or service provider. Edge devices include medical registration machines, printers, financial self-service machines, tablets and other locally deployed devices, etc. For details, see https: / / zhuanlan.zhihu.com / p / 510554675.

[0003] With the improvement of chip technology and computing power, artificial intelligence edge devices carry more and more production and life information and personal privacy data. Edge computing is another new computing model after distributed computing, grid computing, and cloud computing. Edge computing takes cloud computing as the core, modern communication networks as the means, and massive intelligent terminals as the frontier. By optimizing resource allocation, it makes computing, storage, transmission, application and other services more intelligent, and has the ability to coordinate resources with complementary advantages and deep collaboration. It is a new computing model that integrates cloud, network, terminal, and intelligence. Figure 1 Figure 2 shows the physical boundaries of edge computing.

[0004] The transmission, analysis, processing and storage of big data poses a huge challenge to network bandwidth, and the birth of edge computing provides the necessary technical means to solve this problem. A large number of massive terminal devices are the first stop for data collection, and transmitting a large amount of data to the cloud computing center for processing will lead to untimely data processing and even cause system accidents, especially in application scenarios such as industrial Internet, VR / AR, etc. that require high reliability and real-time performance.

[0005] According to the IDC report (Internet Data Center), 80% of the data is unstructured data (Unstructured Data is data with irregular or incomplete structure, no predefined data model, and data that is not convenient to be represented by a two-dimensional logical table in a database). It is estimated that by 2025, the total amount of global data will reach 175ZB. The booming development of the Internet of Things has led to an exponential growth in terminal devices, which has undoubtedly further exacerbated this trend. If the big data processing method still relies on cloud computing and supercomputing models, it is undoubtedly unrealistic. Overly centralized processing methods consume a lot of collection, aggregation, and processing costs, and efficiency cannot be guaranteed. At the same time, many processing tasks do not need to be processed centrally, and can be shared at the front end for preprocessing, reducing unnecessary transmission and storage resource consumption. Therefore, edge computing will reduce the processing pressure of cloud computing models and share a large number of front-end data processing tasks.

[0006] Cloud computing and supercomputing need to rely on computing centers and data centers to provide support for them, and over-concentrated resources consume a lot of electricity. Due to low resource utilization, a large number of resources are idle. The PUE value representing the energy efficiency level is often too high. Foreign data centers are generally less than 2, while most domestic data centers are between 2 and 3. For this reason, the centralized computing model will lead to a huge waste of resources and also bring about excessive power consumption. Edge computing can reduce energy consumption to a certain extent by sharing the storage and computing tasks of the data center without the need to build a large-scale data center.

[0007] The disadvantages of the prior art are:

[0008] Due to the characteristics of edge devices, such as multi-source heterogeneity, openness, and ubiquity, they face more complex security risks and challenges. Edge devices are usually unattended or in unsafe physical environments, with large differences in their own security capabilities, making them easy targets for attackers and becoming a weak link in the overall security of the system.

[0009] How to solve the security issues of edge devices is the key to further promote the rapid development of the artificial intelligence industry. From the current development and practice of security technology, based on trusted computing technology, the security protection capabilities of edge devices themselves can be improved, communication security can be enhanced, and the terminal security management capabilities of cloud servers can be improved, thus improving the overall security of artificial intelligence applications in all aspects.

[0010] After years of exploration and practice, my country has formed completely independent and controllable products such as the PKS system in the fields of chips and operating systems, and has been widely used in government affairs, energy, finance, military industry and other fields, becoming the core foundation of my country's independent, secure and green computing information system. The PKS system is a security system composed of Phytium CPU, Kylin operating system and three-dimensional security chain (S-"Security"). The "PKS" system replaces the Wintel (i.e. Windows-Intel architecture) system in the core areas of the country and some industries. In the future, as the bidding objects sink and the industry application scenarios open up, the advantages will be magnified and the market share will be further expanded.

[0011] AI edge computing combines AI technology with edge computing technology to run AI algorithms on devices capable of edge computing. The benefit is that data can be processed in real time without having to connect to a cloud platform. Some common use cases for edge AI include self-driving cars, drones, facial recognition, and digital assistants. In response to trusted security issues in the field of AI edge computing, we can focus on the localization of the AI ​​technology security framework and the security authentication control of edge devices based on PKS system security and trusted computing technology. With the support of the native security capabilities and mature ecology of the localized PKS system, we can ensure the autonomy, controllability, and security of AI application data. Summary of the invention

[0012] The technical problem to be solved by the present invention is to provide a method for the security control of artificial intelligence edge devices based on the PKS system. The existing PKS trusted computing is only the underlying hardware computing resource and is not combined with artificial intelligence or authentication authorization. The R&D personnel use the hardware foundation of the PKS system to design and implement interface functions to realize the unified framework of artificial intelligence trusted computing based on the PKS system. Then, in the unified framework of artificial intelligence trusted computing based on the PKS system, the support and design of embedded chips and operating systems are studied. Based on the above unified framework, a unified artificial intelligence edge device security computing trusted framework and control method, as well as a decentralized trusted authentication method, are realized in collaboration with the edge and server ends (including remote cloud services and fog computing for the Internet of Things).

[0013] In order to solve the above problems, the technical solution adopted by the present invention is:

[0014] A unified framework for artificial intelligence trusted computing based on the PKS system, including a function management module and a kernel trusted computing interface, wherein the function management module is located at the middleware layer and the kernel trusted computing interface is located at the operating system layer;

[0015] The functional management module includes password management module, certificate management module, encryption and decryption management module, authentication and authorization management module, and trusted traceability management module;

[0016] Password management module, used to manage the passwords of both communicating parties;

[0017] The certificate management module is used to manage the digital certificates of both communicating parties;

[0018] The encryption and decryption management module is used to encrypt and decrypt the data of the communicating parties;

[0019] Authentication and authorization management module, used to manage the permissions of both communicating parties;

[0020] Trusted traceability management module, used to conduct traceability query on the data involved in the communication between the two parties;

[0021] The kernel trusted computing interface includes software TEE environment interface, compatible conversion interface, sensitive business interface, privacy computing interface, and artificial intelligence behavior security detection interface;

[0022] Software TEE environment interface, used to access the trusted execution environment;

[0023] Regarding how to access the trusted execution environment, you can use the software TEE environment interface, but some will be built into the underlying CPU or operating system, in which case the software TEE environment interface is not required. Therefore, the software TEE environment interface is optional.

[0024] A compatible conversion interface is used to convert incompatible data between the middleware layer and the operating system layer;

[0025] Sensitive service interface, used to access sensitive services;

[0026] Privacy computing interface, used to access privacy computing;

[0027] Artificial intelligence behavior security detection interface, used to perform security detection on artificial intelligence behavior.

[0028] A PKS-based AI edge device secure computing trusted framework, utilizing the AI ​​trusted computing unified framework, including AI trusted computing unified frameworks placed in edge devices and servers respectively;

[0029] The AI ​​trusted computing unified framework placed in the edge device is named the embedded environment AI trusted computing unified framework, which works in conjunction with the trusted computing hardware platform in the edge device.

[0030] The AI ​​trusted computing unified framework placed in the server is named the server environment cloud fog end AI trusted computing unified framework, which works in conjunction with the trusted computing hardware platform in the server;

[0031] Information connection between the unified framework of trusted computing of artificial intelligence in embedded environment and the unified framework of trusted computing of artificial intelligence in cloud and fog end of server environment.

[0032] A PKS-based artificial intelligence edge device security control method utilizes the above-mentioned PKS-based artificial intelligence edge device security computing trusted framework, including the following steps:

[0033] Step 1: Build a trusted framework for secure computing of AI edge devices based on the PKS system, and coordinate with the trusted computing hardware platform in edge devices and servers;

[0034] Step 2: Based on the PKS-based AI edge device secure computing trusted framework, the server implements edge device access, authorization, authentication, and access control management through a complete authorization and authentication process, thereby achieving trusted authentication between edge devices and servers.

[0035] Step 3: After the trusted authentication between the edge device and the server is passed, business data is securely transmitted to realize the secure application of artificial intelligence.

[0036] Furthermore, in step 2, the trusted authentication between the edge device and the server is realized (it can also be called "realizing the trusted authentication of the edge device", which means the same thing, because the authentication is basically the edge device actively accessing the server), which specifically includes the following steps:

[0037] S21, the edge device generates a unique device identity identifier;

[0038] S22. The edge device implements identity authentication of the server through public key exchange: the edge device obtains the public key of the server and applies for the signature information of the server;

[0039] S23. After receiving the signature information request from the edge device in step S2, the server signs the server identity information using its own private key;

[0040] S24. After the edge device obtains the signed server identity information, it uses the corresponding public key to verify the signed server identity information to complete the server identity verification / authentication;

[0041] S25. After the server identity verification / authentication is passed, the edge device selects an encryption algorithm supported by the server based on its own computing power for subsequent password calculations;

[0042] S26. The server performs entity authentication on the edge device: the edge device uses its own public key to encrypt the device identity unique identifier and sends it to the server;

[0043] S27, the server performs device authentication authorization management on the encrypted device identity unique identifier sent, confirms the device identity unique identifier, and issues an edge device authentication token after confirming that it is correct;

[0044] S28. The edge device saves the edge device authentication token and executes the business function.

[0045] Further, in step S21, the edge device generates a unique device identity identifier based on the device elements through encrypted HASH; the device elements include but are not limited to one or more of the device manufacturer name, manufacturer ID, product serial number, chip code and communication identifier. The communication identifier here may include the IMEI number, MAC address, etc. of the communication module, but is not limited to this.

[0046] Further, in step S22, the edge device obtains the public key of the server in real time online or offline;

[0047] Offline method: The server writes the server's public key to the edge device through the edge device management process and saves it in the trusted storage of the edge device; when the edge device starts, it checks whether the server's public key exists. If it exists, it directly obtains the server's public key from the trusted storage of the edge device; if it does not exist, it sends a public key request to the server through an online method, and the edge device obtains the server's public key from the server's trusted storage.

[0048] Furthermore, in the secure transmission of business data in step 3, a symmetric key algorithm is used, which specifically includes the following steps:

[0049] Step S31: The edge device checks whether a private symmetric key C for encrypting business data already exists in the local trusted environment and is within the validity period. If not, the edge device requests the server's public key A from the server.

[0050] Step S32: The server obtains the server's public key A from the trusted environment, and returns the server's public key A to the edge device;

[0051] Step S33: The edge device verifies the legitimacy of the public key A, and stores it in a local trusted environment after successful verification; it should be noted here that the trusted environment belongs to the host itself;

[0052] Step S34: The edge device generates a private symmetric key C for business data encryption, encrypts the key C with the public key A using an encryption algorithm, and then uploads the encryption result to the server;

[0053] Step S35: The server uses the private key of public key A to decrypt and obtain key C, stores it in a trusted environment, and sets a validity period;

[0054] Step S36: The server returns the hash value of the key C to the edge device, and the edge device performs a consistency check on the returned hash value. If it fails, it returns to step S34 or terminates.

[0055] Step S37: The edge device subsequently uses the key C to encrypt the service data each time a service request is made, and transmits the encrypted service data to the server.

[0056] Step S38: The server uses the valid key C to decrypt the service data. If the decryption is successful, execute step S39;

[0057] Step S39: The server continues to process the service, encrypts the returned result using the key C, forms a processing result, and returns the processing result to the edge device;

[0058] Step S30: The edge device uses the key C to decrypt the processing result. If the decryption fails, the edge device is notified of the reason for the failure.

[0059] Further, in step S30, the failure reasons include, but are not limited to, one or more of "key C does not exist", "key C does not match", and "expired".

[0060] A decentralized edge device trusted authentication method utilizes the above-mentioned security control method. This method regards the edge devices in the area as edge device clusters. No central server is set in the edge device cluster. Instead, any device is selected as the central device and plays the role of the central server.

[0061] Further, the following steps are included:

[0062] Step 41: Generate a unique device identity identifier for each edge device;

[0063] Step 42, planning leader candidate devices: among the edge devices in the region, according to the set principle, determine some or all edge devices as the leader candidates of the region;

[0064] Step 43: Through the message broadcast mechanism, support the push of messages such as cluster election, election result broadcast, leader device replacement, etc., to achieve status message communication between edge devices;

[0065] Step 44, leader device election at system startup: According to the business scope and processing capacity of the edge device in the region, the edge device status is divided into four states: campaign, follower, reserve and leader. The edge device in the leader state plays the role of the central server. At system startup, each candidate edge device is placed in the campaign state, and the non-candidate edge devices are placed in the reserve state.

[0066] When more than half or a certain number of edge devices are turned on, the election process begins. The state of the elected edge device is set to the leader state, and the other candidate edge devices are changed to the follower state; the edge devices started subsequently are directly in the follower or reserved state;

[0067] Step 45: Leader device election during operation: When the edge device in the current leader state fails, the system conducts a new round of elections;

[0068] The candidate edge device changes to the election state and re-election of the leader device is carried out;

[0069] Step 46: Use the device identity unique identifier to authenticate to the leader node: After receiving the notification message from the leader device, use the device identity unique identifier of the edge device to complete the authentication to the leader device.

[0070] The above process is implemented on the edge device using hardware or software, and decentralized trusted authentication is implemented with the support of a trusted environment.

[0071] Due to the adoption of the above technical solution, the beneficial effects achieved by the present invention are:

[0072] The business data key of the present invention is generated by the edge device on the data production side, avoiding the risk of the server-side generated transmission being stolen.

[0073] The present invention proposes a security authentication control method for artificial intelligence edge devices to achieve trusted and secure storage and unified authentication control, and proposes a decentralized collaboration model and an end-to-end business collaboration model, which can effectively ensure the information security of the data source end of artificial intelligence applications and improve the system's ability to resist attacks and tampering.

[0074] Edge devices are widely used in the fields of artificial intelligence and intelligent manufacturing. The artificial intelligence edge device control method proposed in this invention solves the problem of secure application in embedded devices, supports the integration and improvement of the system capabilities of embedded platforms from chips, operating systems, and secure and trusted computing, and the proposed authentication method and data encryption transmission method ensure the information security of the edge end and enhance the overall robustness of the security of the artificial intelligence cloud edge end. This method has launched successful industry solutions in the financial, government, energy, communications, and medical industries. The technical framework and research results can work together with the PKS system to promote the development and practice of trusted artificial intelligence for various industries.

[0075] The present invention can make full use of the native security mechanism of the PKS system to create a full-link security mode from the upper-layer application to the underlying operating system, so as to achieve application security, communication security, data security, and all-round protection of the entire system. At the same time, it explores the application of various security mechanisms in the AI ​​industry, ensures the safety and reliability of products, effectively prevents the loss of sensitive and private data, and provides a variety of security protection mechanisms for the future application of artificial intelligence in various industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A physical boundary diagram of edge computing involved in the background of the present invention;

[0077] Figure 2 It is a unified framework of artificial intelligence trusted computing based on PKS system of the present invention;

[0078] Figure 3 It is a trusted framework for secure computing of artificial intelligence edge devices based on the PKS system of the present invention;

[0079] Figure 4 This is a flow chart of trusted authentication between edge devices and servers in the present invention;

[0080] Figure 5 This is a diagram of the edge device business data key negotiation and secure transmission process in the present invention. DETAILED DESCRIPTION

[0081] The present invention is further described in detail below with reference to the embodiments.

[0082] The overall idea of ​​the present invention is: edge devices generally adopt embedded operating systems, or are implemented by SOC, and the hardware platform is implemented with low-power embedded chips (SOC, ASIC), etc., and a chip-level secure computing trusted environment is adopted. Through a complete authorization and authentication method, the security authentication of edge intelligent devices is realized, and the security application of artificial intelligence is realized.

[0083] First, develop a unified framework for trusted computing of artificial intelligence based on the PKS system, such as Figure 2 As shown in the box in the upper middle part, Figure 2 The rest of the part is the trusted computing hardware platform and information flow or function definition, which does not belong to the protection scope of the present invention. It includes a function management module and a kernel trusted computing interface, the function management module is located in the middleware layer, and the kernel trusted computing interface is located in the operating system layer;

[0084] The functional management module includes password management module, certificate management module, encryption and decryption management module, authentication and authorization management module, and trusted traceability management module;

[0085] Password management module, used to manage the passwords of both communicating parties;

[0086] The certificate management module is used to manage the digital certificates of both communicating parties;

[0087] The encryption and decryption management module is used to encrypt and decrypt the data of the communicating parties;

[0088] Authentication and authorization management module, used to manage the permissions of both communicating parties;

[0089] Trusted traceability management module, used to conduct traceability query on the data involved in the communication between the two parties;

[0090] The kernel trusted computing interface includes software TEE environment interface, compatible conversion interface, sensitive business interface, privacy computing interface, and artificial intelligence behavior security detection interface;

[0091] Software TEE environment interface, used to access the trusted execution environment;

[0092] It is necessary to make a special statement here: Regarding how to access the trusted execution environment, the software TEE environment interface can be used, but some will be built into the underlying CPU or operating system, in which case the software TEE environment interface is not required. Therefore, the TEE environment interface can be used as an alternative or optional option. Figure 2 The software TEE environment interface is drawn.

[0093] A compatible conversion interface is used to convert incompatible data between the middleware layer and the operating system layer;

[0094] Sensitive service interface, used to access sensitive services;

[0095] Privacy computing interface, used to access privacy computing;

[0096] Artificial intelligence behavior security detection interface, used to perform security detection on artificial intelligence behavior.

[0097] Then, in the unified framework of AI trusted computing based on PKS system, the support and design of embedded chips and operating systems are studied. Based on this unified framework of AI trusted computing, a unified AI edge device secure computing trusted framework based on PKS system is realized with the collaboration of edge and server (including remote cloud services and fog computing for IoT), such as Figure 3 The PKS-based AI edge device secure computing trusted framework includes an AI trusted computing unified framework placed in edge devices and servers respectively;

[0098] The AI ​​trusted computing unified framework placed in the edge device is named the embedded environment AI trusted computing unified framework, which works in conjunction with the trusted computing hardware platform in the edge device.

[0099] The AI ​​trusted computing unified framework placed in the server is named the server environment cloud fog end AI trusted computing unified framework, which works in conjunction with the trusted computing hardware platform in the server;

[0100] Information connection between the unified framework of trusted computing of artificial intelligence in embedded environment and the unified framework of trusted computing of artificial intelligence in cloud and fog end of server environment.

[0101] In this way, with the support of a trusted computing platform, starting from the edge device, based on a secure root of trust, using cryptographic principles, a trust chain is established from the secure root of trust to verify the security of hardware platforms, operating systems, application software, etc., and to build a secure and trusted computing environment. The system can also perform trusted management of edge devices, monitor and evaluate abnormal behaviors, trace and block malicious attacks, etc.

[0102] With the support of the edge device secure computing trusted framework, the industrial-grade authentication standards are optimized and designed. With the support of the cloud / fog terminal, the trusted authentication of edge devices is realized, the authenticity of the devices is identified, and the artificial intelligence services are authorized. The cloud / fog terminal realizes the registration, certificate management, cancellation and authentication access of the devices.

[0103] like Figure 3 As shown in the two boxes on the left and right in the upper middle part, the unified framework of artificial intelligence trusted computing in the embedded environment and the unified framework of artificial intelligence trusted computing in the cloud and fog end of the server environment are the unique design structures of the present invention. Figure 2 The rest of the content is either a trusted computing hardware platform and information flow or function definition, which does not fall within the scope of protection of the present invention. Or it is based on the unique design structure of the present invention to carry out special supporting services.

[0104] Continue as Figure 3 As shown in the figure, in the system framework, according to the hierarchical design, the following are implemented respectively:

[0105] 1. Trusted storage at the edge

[0106] It provides secure storage capabilities for device unique identification and certificates, trusted roots, server public keys, exchange data keys, and privacy data, and supports secure memory.

[0107] 2. Trusted collaboration between cloud and end at the operating system level

[0108] The underlying automatic discovery, automatic networking, key and certificate exchange functions required by artificial intelligence are implemented in the operating system kernel, relying on IoT technology and cryptographic technology to build a trust chain and improve the collaborative capabilities of the cloud, edge and end. The trusted collaboration of the cloud / end at the operating system level accelerates the calculation of "trusted authentication of edge devices and secure transmission of business data", making full use of the computing power of each device for computing power collaboration.

[0109] In addition, it is also possible to design a computing model that supports decentralization (the decentralized working ability of equipment, ensuring the uninterrupted provision of services, and the business continuity capabilities between devices and local fog terminals).

[0110] 3. Intelligent business collaboration in a unified trusted computing framework

[0111] Provide business collaboration for artificial intelligence services, realize algorithm collaboration (cloud and edge implement their own algorithms, and the distributed capabilities of algorithms, such as multi-terminal cluster services, multiple devices collaborate to participate in artificial intelligence identity recognition, fire identification, etc.), computing power collaboration (collaborative cloud, edge and terminal intelligent computing power), and business collaboration (business applications under privacy data protection, authentication and authorization collaboration, etc.).

[0112] 4. Trusted authentication of edge devices

[0113] The server implements the access, authorization, authentication and access control management of edge devices, and realizes the trusted authentication of devices.

[0114] 5. Secure transmission of business data

[0115] Provide secure transmission capabilities for important AI business and important privacy data, and achieve one-machine-one-secret secure communication.

[0116] Functionally, based on the edge device secure computing trusted framework, it can achieve:

[0117] 1) Prevent illegal devices from accessing;

[0118] 2) Prevent unauthorized access;

[0119] 3) Prevent AI data from being tampered with during transmission.

[0120] The framework design complies with the Device Identity Composition Engine (DICE) specification of TCG (Trusted Computing Group), supports environments with or without hardware TPM, implements various device / software functions in layers, and each layer has an independent security protection design.

[0121] The present invention also discloses a PKS-based artificial intelligence edge device security control method, which utilizes the above-mentioned PKS-based artificial intelligence edge device security computing trusted framework, and includes the following steps:

[0122] Step 1: Build a trusted framework for secure computing of AI edge devices based on the PKS system, and coordinate with the trusted computing hardware platform in edge devices and servers;

[0123] Step 2: According to the PKS-based AI edge device secure computing trusted framework, through a complete authorization and authentication process, the server implements edge device access, authorization, authentication, and access control management, thereby achieving trusted authentication between edge devices and servers; trusted authentication between edge devices and servers can also be called "trusted authentication of edge devices", which means the same thing, because authentication is basically the edge device actively accessing the server.

[0124] Step 3: After the trusted authentication between the edge device and the server is passed, business data is securely transmitted to realize the secure application of artificial intelligence.

[0125] Specifically, in step 2, the trusted authentication between the edge device and the server is implemented, which includes the following steps:

[0126] S21, the edge device generates a unique device identity identifier;

[0127] S22. The edge device implements identity authentication of the server through public key exchange: the edge device obtains the public key of the server and applies for the signature information of the server;

[0128] S23. After receiving the signature information request from the edge device in step S2, the server signs the server identity information using its own private key;

[0129] S24. After the edge device obtains the signed server identity information, it uses the corresponding public key to verify the signed server identity information to complete the server identity verification / authentication;

[0130] S25. After the server identity verification / authentication is passed, the edge device selects an encryption algorithm supported by the server based on its own computing power for subsequent password calculations;

[0131] S26. The server performs entity authentication on the edge device: the edge device uses its own public key to encrypt the device identity unique identifier and sends it to the server;

[0132] S27, the server performs device authentication authorization management on the encrypted device identity unique identifier sent, confirms the device identity unique identifier, and issues an edge device authentication token after confirming that it is correct;

[0133] S28. The edge device saves the edge device authentication token and executes the business function.

[0134] Then, in step S21, the edge device generates a unique device identity identifier based on the device elements through encrypted HASH; the device elements include but are not limited to one or more of the device manufacturer name, manufacturer ID, product serial number, chip code and communication identifier. The communication identifier here may include the IMEI number and MAC address of the communication module, but is not limited thereto.

[0135] Then, in step S22, the edge device obtains the public key of the server in real time online or offline;

[0136] Offline method: The server writes the server's public key to the edge device through the edge device management process and saves it in the trusted storage of the edge device; when the edge device starts, it checks whether the server's public key exists. If it exists, it directly obtains the server's public key from the trusted storage of the edge device; if it does not exist, it sends a public key request to the server through an online method, and the edge device obtains the server's public key from the server's trusted storage.

[0137] Next, in the secure transmission of business data in step 3, a symmetric key algorithm is used, combined with Figure 5 , specifically including the following steps:

[0138] (1) The edge device checks whether there is a private symmetric key C for business data encryption in the local trusted environment and whether it is within the validity period. If not, it requests the server's public key A from the server;

[0139] (2) The server obtains the server's public key A from the trusted environment and returns the server's public key A to the edge device;

[0140] (3) The edge device verifies the legitimacy of public key A. After successful verification, it is stored in the local trusted environment, because the trusted environment belongs to the host itself;

[0141] (4) The edge device generates a private symmetric key C for business data encryption, encrypts the key C with the public key A using an encryption algorithm, and then uploads the encrypted result to the server;

[0142] (5) The server uses the private key of public key A to decrypt and obtain key C, stores it in a trusted environment, and sets a validity period;

[0143] (6) The server returns the hash value of key C to the edge device, and the edge device performs a consistency check on the returned hash value. If it fails, it returns to step S34 or terminates;

[0144] (7) Each subsequent edge device uses key C to encrypt business data and transmit it to the server when a business request is made;

[0145] (8) The server uses the valid key C to decrypt the business data. If the decryption is successful, proceed to step (9).

[0146] (9) The server continues to process the business, encrypts the return result using key C, forms a processing result, and returns the processing result to the edge device;

[0147] (10) The edge device uses key C to decrypt the processing result. If the decryption fails, the edge device is notified of the failure reason. The failure reason includes but is not limited to one or more of "key C does not exist", "key C does not match", and "expired".

[0148] The process safety analysis of (1)-(10) above is as follows:

[0149] In (1)-(2), it is assumed that there is a middleman who can obtain the certificate public key A, but cannot modify it because the edge device will verify the legitimacy of the certificate and the middleman cannot forge it, ensuring the trustworthiness of the server.

[0150] In (4)-(6), it is assumed that there is a middleman who can forge a key C (called C1) and encrypt it using the public key A. The server stores C1.

[0151] In (7)-(8), the edge device uses C to encrypt data, and the server uses C1 to decrypt, resulting in decryption failure. At this time, the server informs the edge device that the keys are inconsistent, ensuring that the encryption key cannot be forged.

[0152] The following is a further description with reference to the accompanying drawings.

[0153] The process of the artificial intelligence edge device security control method based on the PKS system of the present invention is as follows: Figure 4 As shown, combined Figure 4 For example:

[0154] 2.3.2.1 Equipment Identification

[0155] The edge device generates a unique identifier for the device identity based on the following factors through encrypted hashing:

[0156] 1) Manufacturer name or ID;

[0157] 2) Product serial number;

[0158] 3) Chip coding;

[0159] 4) IMEI number, MAC address, etc. of the communication module.

[0160] This identifier is uniquely guaranteed through cryptographic technology and is used for device authentication management and authorization.

[0161] 2.3.2.2 Server Authentication and Encryption Algorithm Negotiation

[0162] In order to protect the data security on edge devices and prevent the devices from being illegally remotely controlled, a general asymmetric encryption or certificateless encryption technology (such as PKI / CLA, to achieve zero-trust secure access) authentication method is adopted to realize the edge device's identity authentication to the server through the "public key exchange" method.

[0163] The edge device first obtains the public key of the server. To support system deployment and dynamic migration, it supports both offline and online methods of obtaining public keys. Offline, through the device management process, the server public key is written to the edge device and saved in the trusted storage. When the edge device starts, it checks whether the public key exists. If not, it sends a request to the server to obtain the public key through the online method.

[0164] The server uses the private key to sign the server identity information (including server identity, encryption algorithm capability list, etc.). The edge device uses the corresponding public key to verify the signature and authenticate the server. It can also use a certificate system similar to X.509 to verify the server identity through a certificate chain.

[0165] According to the computing power of the edge device itself, select the encryption algorithm supported by the server for subsequent password calculations.

[0166] 2.3.2.3 Server Entity Authentication of Edge Devices

[0167] The edge device uses the public key to encrypt the device identity and sends it to the server. The server's device authentication and authorization management function uses blacklist, whitelist, access control and other technologies to confirm the device identity and issue a device authentication token.

[0168] When the edge device reconnects or times out, the token becomes invalid and the edge device re-initiates the authentication process.

[0169] In the process of key negotiation and secure transmission of business data of edge devices, in addition to the TLS (Transport Layer Security) communication protocol encryption method, for the business data of artificial intelligence, based on the trusted authentication of edge devices, the data transmission key is agreed with the server through the key negotiation mechanism (the key has a validity period and automatically becomes invalid after expiration), and a symmetric key algorithm with less computational effort is used to ensure the security of business data transmission and computing efficiency.

[0170] In order to improve processing efficiency, we can consider two aspects and adopt the following two methods:

[0171] 1) Hardware encryption calculation: The hardware module of the edge device has a built-in encryption calculation unit to increase the speed of encryption and decryption. The trusted framework provides an API interface to the upper-layer application;

[0172] 2) Based on the trusted authentication of edge devices, and within the scope permitted by the trusted computing policy, lightweight encryption methods and key lengths are used for secure transmission of business data.

[0173] The above two methods are not related, they just accelerate encryption and decryption from two aspects, one is to use hardware resources to accelerate, and the other is to use a simple encryption algorithm. Figure 5 (7)-(10) are used for business data transmission.

[0174] The following process is designed to be performed after the edge device has completed trusted authentication.

[0175] The design feature of the secure transmission of business data is that the business data key is generated by the edge device on the data production side, avoiding the risk of the transmission generated by the server being stolen.

[0176] This design does not limit the cryptographic algorithm. You can use the device authentication phase to negotiate and determine the algorithm from the encryption algorithms supported by the server. Add an algorithm agreement field in the message to indicate the negotiated algorithm, or implement automatic adaptation of the cryptographic algorithm. In this way, the edge device and server implement various algorithms including national encryption. The specific process is as follows Figure 5 .

[0177] The edge device checks whether the private symmetric key C for business data encryption exists in the local trusted environment and is valid. If not, it requests the public key A of the server certificate from the server.

[0178] • The edge device checks the legitimacy of public key A and stores it in a local trusted environment after successful verification.

[0179] • The edge device generates a private symmetric key C, encrypts key C with public key A using an encryption algorithm, and then uploads the encrypted result to the server.

[0180] • The server uses the private key of certificate A to decrypt and store key C in a trusted environment and sets a validity period. The server returns the hash value of key C to the edge device, which performs a consistency check.

[0181] • The edge device subsequently uses key C to encrypt and transmit to the server each time a service request is made.

[0182] • The server uses the valid key C to decrypt the business data. If the decryption is successful, the business processing continues. If the decryption fails, the edge device is notified of the failure reason, such as "key C does not exist", "key C does not match", "expired", etc.

[0183] 2.3.4 Decentralized Trusted Authentication

[0184] At the same time, the present invention also discloses another decentralized edge device trusted authentication method, which regards the edge devices in the area as an edge device cluster. No central server is set in the edge device cluster, but any device is selected as the central device and plays the role of the central server.

[0185] In applications with high reliability requirements, a certain geographical area or business area needs to maintain business sustainability without a central server. In this case, decentralized trusted authentication needs to be supported when edge devices are connected (decentralization at the business level is not included in this solution design).

[0186] At present, there are decentralized terminal security authentication systems based on blockchain and DID (Decentralized Identifiers) technology, but blockchain technology is relatively complex and not easy to implement. We regard the edge devices in the region as edge device clusters and propose to use an optimized cluster election system to implement a lightweight decentralized edge device trusted authentication method.

[0187] 2.3.4.1 Use the device identification method in the edge device trusted authentication method to generate an edge device identification;

[0188] 2.3.4.2 Planning leader candidate devices

[0189] Among the edge devices in the region, some or all edge devices are determined as the leader candidates of the region according to certain principles.

[0190] Determine candidate devices and their weights based on the following parameters:

[0191] Embedded chip computing power;

[0192] Embedded chip concurrency capabilities;

[0193] Embedded chip memory;

[0194] Communication IO capabilities of edge devices;

[0195] Network environment and communication routing, etc.

[0196] 2.3.4.3 Implementing the message communication mechanism of edge device cluster

[0197] Through the message broadcast mechanism, it supports the push of messages such as cluster election, election result broadcast, and leader device replacement, and realizes status message communication between edge devices.

[0198] 2.3.4.4 Leader device election at system startup

[0199] The leader election method provided by the present invention is selected according to the business scope and processing capability of the edge device in the region.

[0200] The leader election method divides the device status into four states: campaign, follower, reserve and leader. When the system starts, each candidate edge device is placed in the campaign state, and non-candidate edge devices are placed in the reserve state.

[0201] When more than half or a certain number of edge devices are turned on, the election process begins. The state of the elected edge device is set to the leader state, and the other candidate devices are changed to the follower state.

[0202] The edge devices started subsequently are directly in the follower or reserved state.

[0203] 2.3.4.5 Leader device election during operation

[0204] When the current leader device fails, the system conducts a new round of elections.

[0205] The candidate device changes to the election state and re-election of the leader device is carried out.

[0206] 2.3.4.6 Use edge device identification to authenticate to the leader node

[0207] After receiving the notification message from the leader device, the edge device uses the edge device identifier to complete the authentication with the leader device.

[0208] The above process is implemented on the edge device using hardware or software, and decentralized trusted authentication is implemented with the support of a trusted environment.

[0209] The above leader election method only exists in the current election method. Other leader election methods may be different. For example, you can choose the distributed election algorithm raft and select the appropriate method according to the specific situation.

[0210] The present invention discloses a unified framework for artificial intelligence trusted computing based on the PKS system, a trusted framework for secure computing of artificial intelligence edge devices based on the unified framework for artificial intelligence trusted computing, a security control method for artificial intelligence edge devices based on the trusted framework for secure computing of artificial intelligence edge devices, and a trusted authentication method for decentralized edge devices. The secure authentication control method for artificial intelligence edge devices based on the PKS system realizes trusted secure storage and unified authentication control, proposes a decentralized collaborative mode and an end-to-end business collaborative mode, which can effectively ensure the information security of the data source end of artificial intelligence applications and improve the system's ability to resist attacks and tampering.

[0211] The present invention can make full use of the native security mechanism of the PKS system to create a full-link security mode from the upper-layer application to the underlying operating system, so as to achieve application security, communication security, data security, and all-round protection of the entire system. At the same time, it explores the application of various security mechanisms in the AI ​​industry, ensures the safety and reliability of products, effectively prevents the loss of sensitive and private data, and provides a variety of security protection mechanisms for the future application of artificial intelligence in various industries.

Claims

1. A PKS-based artificial intelligence edge device security control method, which utilizes a PKS-based artificial intelligence edge device security computing trusted framework, characterized in that: The steps include: Step 1: Build a trusted framework for secure computing of AI edge devices based on the PKS system, and coordinate with the trusted computing hardware platform in edge devices and servers; Step 2: Based on the PKS-based AI edge device secure computing trusted framework, the server implements edge device access, authorization, authentication, and access control management through a complete authorization and authentication process, thereby achieving trusted authentication between edge devices and servers. Step 3: After the trusted authentication between the edge device and the server is passed, the business data is securely transmitted to realize the secure application of artificial intelligence; The PKS-based AI trusted computing unified framework includes a function management module and a kernel trusted computing interface. The function management module is located at the middleware layer, and the kernel trusted computing interface is located at the operating system layer. The functional management module includes password management module, certificate management module, encryption and decryption management module, authentication and authorization management module, and trusted traceability management module; Password management module, used to manage the passwords of both communicating parties; The certificate management module is used to manage the digital certificates of both communicating parties; The encryption and decryption management module is used to encrypt and decrypt the data of the communicating parties; Authentication and authorization management module, used to manage the permissions of both communicating parties; Trusted traceability management module, used to conduct traceability query on the data involved in the communication between the two parties; The kernel trusted computing interface includes software TEE environment interface, compatible conversion interface, sensitive business interface, privacy computing interface, and artificial intelligence behavior security detection interface; Software TEE environment interface, used to access the trusted execution environment; A compatible conversion interface is used to convert incompatible data between the middleware layer and the operating system layer; Sensitive service interface, used to access sensitive services; Privacy computing interface, used to access privacy computing; Artificial intelligence behavior security detection interface, used to perform security detection on artificial intelligence behavior; It also includes a unified framework for trusted computing of artificial intelligence, which is placed in edge devices and servers respectively; The AI ​​trusted computing unified framework placed in the edge device is named the embedded environment AI trusted computing unified framework, which works in conjunction with the trusted computing hardware platform in the edge device. The AI ​​trusted computing unified framework placed in the server is named the server environment cloud fog end AI trusted computing unified framework, which works in conjunction with the trusted computing hardware platform in the server; Information connection between the unified framework of trusted computing of artificial intelligence in embedded environment and the unified framework of trusted computing of artificial intelligence in cloud and fog end of server environment.

2. According to claim 1, a PKS system-based artificial intelligence edge device security control method is characterized in that: In step 2, the trusted authentication between the edge device and the server is implemented, which specifically includes the following steps: S21, the edge device generates a unique device identity identifier; S22. The edge device implements identity authentication of the server through public key exchange: the edge device obtains the public key of the server and applies for the signature information of the server; S23. After receiving the signature information request from the edge device in step S2, the server signs the server identity information using its own private key; S24. After the edge device obtains the signed server identity information, it uses the corresponding public key to verify the signed server identity information to complete the server identity verification / authentication; S25. After the server identity verification / authentication is passed, the edge device selects an encryption algorithm supported by the server based on its own computing power for subsequent password calculations; S26. The server performs entity authentication on the edge device: the edge device uses its own public key to encrypt the device identity unique identifier and sends it to the server; S27, the server performs device authentication authorization management on the encrypted device identity unique identifier sent, confirms the device identity unique identifier, and issues an edge device authentication token after confirming that it is correct; S28. The edge device saves the edge device authentication token and executes the business function.

3. According to claim 2, a PKS system-based artificial intelligence edge device security control method is characterized in that: In step S21, the edge device generates a unique device identity identifier based on device elements; the device elements include but are not limited to one or more of the device manufacturer name, manufacturer ID, product serial number, chip code and communication identification.

4. According to claim 2, a method for security control of artificial intelligence edge devices based on PKS system is characterized in that: In step S22, the edge device obtains the public key of the server in real time online or offline; Offline method: The server writes the server's public key to the edge device through the edge device management process and saves it in the trusted storage of the edge device; When the edge device starts, it checks whether the server's public key exists. If it exists, it directly obtains the server's public key from the trusted storage of the edge device; If it does not exist, a request for obtaining the public key is sent to the server online, and the edge device obtains the server's public key from the server's trusted storage.

5. According to claim 1, a method for security control of artificial intelligence edge devices based on PKS system is characterized in that: In the secure transmission of business data in step 3, a symmetric key algorithm is used, which specifically includes the following steps: Step S31: The edge device checks whether a private symmetric key C for encrypting business data already exists in the local trusted environment and is within the validity period. If not, the edge device requests the server's public key A from the server. Step S32: The server obtains the server's public key A from the trusted environment, and returns the server's public key A to the edge device; Step S33: The edge device verifies the legitimacy of public key A and stores it in a local trusted environment after successful verification. Step S34: The edge device generates a private symmetric key C for business data encryption, encrypts the key C with the public key A using an encryption algorithm, and then uploads the encryption result to the server; Step S35: The server uses the private key of public key A to decrypt and obtain key C, stores it in a trusted environment, and sets a validity period; Step S36: The server returns the hash value of the key C to the edge device, and the edge device performs a consistency check on the returned hash value. If it fails, it returns to step S34 or terminates. Step S37: The edge device subsequently uses the key C to encrypt the service data each time a service request is made, and transmits the encrypted service data to the server. Step S38: The server uses the valid key C to decrypt the service data. If the decryption is successful, execute step S39; Step S39: The server continues to process the service, encrypts the returned result using the key C, forms a processing result, and returns the processing result to the edge device; Step S30: The edge device uses the key C to decrypt the processing result. If the decryption fails, the edge device is notified of the reason for the failure.

6. According to claim 5, a method for security control of artificial intelligence edge devices based on PKS system is characterized in that: In step S30, the failure reasons include, but are not limited to, one or more of "key C does not exist", "key C does not match", and "expired".

7. A decentralized edge device trusted authentication method, utilizing the security control method according to any one of claims 1 to 6, characterized in that: The edge devices in the area are regarded as edge device clusters. There is no central server in the edge device cluster. Instead, any device is selected as the central device and plays the role of the central server.

8. A decentralized edge device trusted authentication method according to claim 7, characterized in that: The steps include: Step 41: Generate a unique device identity identifier for each edge device; Step 42, planning leader candidate devices: among the edge devices in the region, according to the set principle, determine some or all edge devices as the leader candidates of the region; Step 43: Through the message broadcast mechanism, cluster election, election result broadcast, and leader device replacement message push are supported to realize status message communication between edge devices; Step 44, leader device election at system startup: select a suitable leader election method according to the business scope and processing capacity of the edge device in the region, and divide the edge device status into four states: campaign, follower, reserve and leader. The edge device in the leader state plays the role of the central server. At system startup, each candidate edge device is placed in the campaign state, and non-candidate edge devices are placed in the reserve state. When more than half or a certain number of edge devices are turned on, the election process begins. The state of the elected edge device is set to the leader state, and the other candidate edge devices are changed to the follower state; the edge devices started subsequently are directly in the follower or reserved state; Step 45: Leader device election during operation: When the edge device in the current leader state fails, the system conducts a new round of elections; The candidate edge device changes to the election state and re-election of the leader device is carried out; Step 46: Use the device identity unique identifier to authenticate to the leader node: After receiving the notification message from the leader device, use the device identity unique identifier of the edge device to complete the authentication to the leader device.

Images