AI digital identity hosting gateway system and method based on trusted execution environment
By leveraging the Trusted Execution Environment (TEE)-based AI digital identity hosting gateway system, which combines TEE hardware security capabilities with distributed identity protocols, the system addresses the issues of privacy protection for intermediate tensors in AI models and cross-device invocation. It enables secure hosting and privacy computing on consumer-grade hardware, providing a continuous and consistent service experience across devices.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN TONGKANG CHUANGZHI TECH CO LTD
- Filing Date
- 2026-02-07
- Publication Date
- 2026-05-15
AI Technical Summary
Existing technologies cannot effectively protect the privacy of intermediate tensors in AI models, making cross-device invocation difficult. Furthermore, traditional privacy computing frameworks cannot achieve real-time inference on low-power devices, leading to loss of data sovereignty and service fragmentation.
An AI digital identity hosting gateway system based on a Trusted Execution Environment (TEE) is adopted, which combines TEE hardware security capabilities with a distributed identity protocol to achieve secure roaming and privacy computing across devices. It provides privacy computing services by securely storing and computing data in a hardware-level trusted execution environment and using verifiable credentials to carry inference context information.
It enables secure hosting of AI digital identities on consumer-grade hardware, ensuring data privacy, providing a continuous and consistent service experience across devices, and enabling low-cost privacy computing on low-power devices.
Smart Images
Figure CN122053155A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence technology, specifically to an AI digital identity hosting gateway system and method based on a trusted execution environment. Background Technology
[0002] Currently, the storage and management of personal digital assets, especially AI digital identities (such as personalized large model parameters, conversation history, and behavioral preferences), faces a dilemma. On the one hand, public cloud services offer convenient synchronization and access, but this leads to a complete loss of user data sovereignty, resulting in continuous privacy risks and subscription fees. On the other hand, while purely local storage solutions (such as mobile phones and NAS) offer data privacy, they suffer from issues such as fragmented user experience, difficulties in cross-device access, and insufficient security.
[0003] Existing technical solutions have the following inherent drawbacks: 1. Long-term hosting and inference process of AI models lack visibility throughout: Traditional technologies can only protect the model file itself, not the intermediate tensors in the inference process. These intermediate vectors can be reverse-engineered to infer privacy and model parameters. 2. Cross-device private AI service continuity issues: Existing verifiable credentials (VCs) are only used for identity authentication and cannot carry inference context information, thus failing to securely transfer the AI dialogue state. 3. Achieving low-cost privacy computing on consumer-grade hardware: Traditional privacy computing frameworks (FHE / MPC) cannot achieve real-time inference on low-power devices such as home routers, lacking applicable paths. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this paper provides an AI digital identity hosting gateway system and method based on a Trusted Execution Environment (TEE). By combining TEE hardware security capabilities, distributed identity protocols, and personal AI model hosting, it resolves the data sovereignty risks of public cloud solutions and the functional fragmentation issues of local solutions. Ultimately, it establishes a core AI digital identity hub at the home network entrance, belonging to the individual user, supporting secure cross-device roaming, and providing privacy-preserving computing services to external entities.
[0005] This invention is implemented as follows: an AI digital identity hosting gateway system based on a trusted execution environment, comprising: The hardware platform, including the system-on-a-chip (SoC) within it, supports a hardware-level trusted execution environment. The secure firmware system runs on the hardware platform and is divided into a logically isolated trusted execution environment and a rich execution environment. The core service suite includes a digital identity vault service deployed within the trusted execution environment for securely storing users’ personal AI models and biometrics; The local secure AI inference service includes a secure inference engine running within the trusted execution environment, used to load and execute inference computations of the personal AI model in an isolated environment; Distributed identity and secure roaming service is used to generate device identities based on the trusted execution environment, issue verifiable credentials containing inference context information, and coordinate the continuation of AI services across devices; A privacy computing service gateway, deployed in the rich execution environment, is used to receive external encrypted computing tasks and schedule them to the secure inference engine within the trusted execution environment for processing. Management applications run on user terminals and are used for system configuration and interaction.
[0006] As a further aspect of the present invention: the digital identity vault service also includes: A dynamic multi-factor access controller is used to enforce policy checks on access requests based on real-time biometric liveness verification. Model version control and non-rollback mechanism are used to bind the stored model version information to the digital certificate to ensure that the model version can only be upgraded in one direction; A sensitive API whitelist mechanism is used to allow only predefined and signed APIs to access sensitive resources within the vault.
[0007] As a further aspect of the present invention: the local secure AI inference service further includes: a model management and scheduler running in the rich execution environment; wherein, the secure inference engine is configured to: receive encrypted input data from the rich execution environment, decrypt it in the trusted execution environment, use the decrypted personal AI model for inference, and return the output result after encryption; the model parameters and intermediate calculation results during the inference process do not leave the trusted execution environment throughout the entire process.
[0008] As a further aspect of the present invention, the local secure AI inference service also includes a fault recovery and replay attack protection mechanism, which is used to record the step hash chain during the inference process and verify the continuity of the hash chain during interruption recovery to prevent replay attacks.
[0009] As a further aspect of the present invention: the distributed identity and secure roaming service is configured as follows: A decentralized identifier for the device is generated based on hardware information within the trusted execution environment; Issue verifiable credentials for verified external devices that have a short validity period and include an inference context digest, session key derivation parameters, and model version fingerprint. A secure tunnel is established between the credentials and the external device, and an encrypted session context is synchronized.
[0010] As a further aspect of the present invention: the distributed identity and secure roaming service also includes: A cross-device migration failure fallback mechanism is used to generate a new context summary containing only the necessary state information for local recovery when roaming fails; The inference state snapshot differential synchronization mechanism is used to synchronize only the difference data between two consecutive inference states when migrating across devices.
[0011] As a further aspect of the present invention, the system further includes a quality of service guarantee mechanism, used to allocate computing resources to tasks of different priorities, dynamically adjust data transmission strategies, or perform task scheduling in the scenarios of trusted execution environment inference, cross-device roaming, and privacy computing services, so as to ensure bandwidth, latency, and reliability.
[0012] Another objective of this invention is to provide an AI digital identity hosting method based on a trusted execution environment, the method comprising the following steps: Perform secure initialization and identity anchoring: complete secure startup verification in the trusted execution environment of the gateway, generate a unique cryptographic identity for the device, and guide the user to register biometric features and encrypt and store them in the initial personal AI model; Perform local privacy-first AI services: preprocess and encrypt user input in the rich execution environment, and pass the encrypted data into the trusted execution environment; decrypt the data in the trusted execution environment and use the personal AI model for inference, and return the encrypted result to the rich execution environment for output; Perform cross-device secure service roaming: In response to service requests from external terminals, verify the requests and issue one-time access credentials within the trusted execution environment; establish a secure tunnel with the external terminal based on the credentials, and synchronize encrypted AI service contexts to achieve service continuity; Open privacy capabilities: Receive external encryption tasks through the privacy computing service gateway; process the task using an authorized personal AI model within the trusted execution environment; and return the encryption result to the requester.
[0013] As a further aspect of the present invention, the AI service that prioritizes local privacy further includes: if the inference process is interrupted, a fault recovery mechanism is enabled to recalculate the context digest and verify the continuity of the hash chain to prevent replay attacks.
[0014] As a further aspect of the present invention, the cross-device secure service roaming also includes: dynamically adjusting the quality of service policy according to network conditions; if the cross-device migration fails, enabling the fallback mechanism to generate a new context summary; and when synchronizing the service context, using differential synchronization to transmit only the state difference data.
[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention establishes a secure foundation for storage and computing through a trusted execution environment, enabling users' AI digital identities to be securely protected on consumer-grade network devices for the first time. This fundamentally overturns the traditional model where data must be uploaded to the cloud for convenient use. Through distributed identity and credential-based authorization mechanisms, personal AI services revolve around the user rather than the device, providing a truly continuous and consistent service experience. By introducing a privacy computing service gateway, devices can securely release AI computing power and model value to the external ecosystem while ensuring absolute user privacy, laying the foundation for new business models. All technologies in this invention are built on existing industrial-grade hardware and open-source software frameworks, possessing a clear engineering implementation path, providing a complete solution for upgrading smart gateway products from "connection pipes" to "home intelligence and security hubs." Attached Figure Description
[0016] Figure 1 This is a system architecture diagram of an embodiment of the present invention.
[0017] Figure 2 This is a schematic diagram illustrating secure access control for the digital identity vault in an embodiment of the present invention.
[0018] Figure 3 This is a schematic diagram of the data flow for secure AI inference in an embodiment of the present invention.
[0019] Figure 4 This is a schematic diagram of signaling and data flow interaction for secure roaming across devices in an embodiment of the present invention.
[0020] Figure 5 This is a flowchart illustrating the internal processing of external calls to the privacy computing service in an embodiment of the present invention. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0022] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.
[0023] like Figure 1 and Figure 2 As shown, this embodiment of the invention provides an AI digital identity hosting gateway system based on a trusted execution environment, comprising: The hardware platform, including the system-on-a-chip (SoC) within it, supports a hardware-level trusted execution environment. The secure firmware system runs on the hardware platform and is divided into logically isolated Trusted Execution Environment (TEE) and Rich Execution Environment (REE). The core service suite includes a digital identity vault service deployed within the trusted execution environment for securely storing users’ personal AI models and biometrics; The local secure AI inference service includes a secure inference engine running within the trusted execution environment, used to load and execute inference computations of the personal AI model in an isolated environment; Distributed identity and secure roaming service is used to generate device identities based on the trusted execution environment, issue verifiable credentials containing inference context information, and coordinate the continuation of AI services across devices; A privacy computing service gateway, deployed in the rich execution environment, is used to receive external encrypted computing tasks and schedule them to the secure inference engine within the trusted execution environment for processing. Management applications run on user terminals and are used for system configuration and interaction.
[0024] In this embodiment of the invention, the hardware platform includes a router motherboard, a system-on-a-chip (SoC) acting as the main control unit and integrating hardware virtualization or TrustZone technology, high-speed memory, and a network interface module. The SoC's TEE secure world is physically and logically isolated from the ordinary world. A Trusted Execution Environment (TEE) runs a streamlined and secure trusted operating system (such as OP-TEE), responsible for hardware root key management, secure storage, cryptographic operations, secure boot, and integrity measurement. A Rich Execution Environment (REE) runs a general-purpose router operating system (such as one based on OpenWrt), responsible for the network protocol stack, device drivers, and general application services. A digital identity vault service is deployed within the TEE.
[0025] In this embodiment of the invention, the secure encrypted storage area in the digital identity vault service is encrypted using a key derived from the unique root key of the TEE hardware. This key is used to persistently store the user's personal AI model fine-tuning parameters, biometric templates, authorization policy chains, and cross-device session context snapshots. Access to the stored data in the digital identity vault service enforces a multi-factor dynamic access control policy based on real-time biometric liveness verification. This includes model version control and a non-rollback mechanism, as well as a sensitive API whitelist mechanism. Only signed, whitelisted APIs can access sensitive resources. Specifically, the digital identity vault service includes a dynamic multi-factor access controller, a model version control and non-rollback mechanism, and a sensitive API whitelist mechanism. The dynamic multi-factor access controller enforces real-time biometric liveness verification for access requests. In addition to device ownership verification, real-time secondary biometric liveness verification is required for critical operations (such as model export and feature updates). The model version control and non-rollback mechanism binds the stored model version information to the digital certificate to ensure that the model version can only be upgraded in one direction, preventing attackers from bypassing security policies by using old model versions. The sensitive API whitelist mechanism allows only predefined and signed APIs to access sensitive resources in the vault, and all non-whitelisted APIs are rejected or rate-limited.
[0026] In this embodiment of the invention, the secure access control data flow for the digital identity vault is as follows: When an application on the REE side requests access to vault data, the request is forwarded to a secure, trusted application within the TEE via the OP-TEE client API. The dynamic multi-factor access controller within the TEE first verifies the legitimacy of the request source and queries the operation type of the request. For critical operations, the controller uses a secure driver to call the microphone module integrated on the motherboard to collect the user's real-time voiceprint and compares it with the encrypted template stored in the TEE's secure storage area. Only after the dual verification of the device certificate and biometrics is successful does the controller use the TEE hardware root key to decrypt the key encapsulating the data, thereby decrypting and returning the required data. All verification and decryption operations are completed within the TEE, and the key and biometric template are never exposed to the REE side.
[0027] like Figure 3As shown in the embodiment of the present invention, the local secure AI inference service further includes: a model management and scheduler running in the rich execution environment; wherein, the secure inference engine is configured to: receive encrypted input data from the rich execution environment, decrypt it within the trusted execution environment, use the decrypted personal AI model for inference, and return the encrypted output result; the model parameters and intermediate calculation results during the inference process do not leave the trusted execution environment at all. In the local secure AI inference service, the input and output data of the secure inference engine are kept encrypted outside the trusted execution environment, the model parameters are only decrypted and used within this environment, and fault recovery and replay attack protection mechanisms are included. TEE-side secure inference engine: As a trusted application within the TEE, it has the ability to load encrypted models and perform tensor calculations. It receives preprocessed and encrypted input data from the REE side, completes inference in an isolated environment, and outputs encrypted results. It ensures that the original user data, model parameters, and intermediate calculation results do not leave the TEE security boundary at all. The REE-side model management and scheduler is responsible for the version management of the AI model, runs a privacy-insensitive general model on the REE side, and acts as a service proxy and request router for the secure inference engine.
[0028] In this embodiment of the invention, the local secure AI inference service further includes a fault recovery and replay attack protection mechanism, used to record the step hash chain during inference and verify the continuity of the hash chain during interruption recovery to prevent replay attacks. The complete data flow of local secure AI inference is as follows: 1. Input and encryption (REE side): The user's voice is converted into a feature vector by the voice front-end processing module on the REE side. Subsequently, the vector is encrypted by a session key negotiated by the TEE. 2. Secure transmission: The encrypted feature vector is transmitted from the REE to the TEE through the shared memory channel of the OP-TEE. 3. Decryption and inference (TEE side): Within the TEE, the secure inference engine first decrypts the feature vector. Then, it loads the encrypted user-private AI model (such as a LoRA weight file) from the digital identity vault, decrypts it in memory, and performs inference calculations. 4. Encryption and return: The generated text result is re-encrypted by the same session key within the TEE and returned to the REE side through the shared memory channel. 5. Output (REE side): After applying the decryption result, the REE side calls the TTS engine to generate voice feedback to the user. During this process, the user's original voice features, private model parameters, and intermediate inference data are all confined to the physical isolation environment of the TEE.
[0029] like Figure 4As shown in this embodiment of the invention, the distributed identity and secure roaming service is configured to: generate a decentralized identifier for the device based on hardware information within the trusted execution environment, serving as the root identity of the gateway in the identity network; issue verifiable credentials with short validity periods, containing inference context digests, session key derivation parameters, and model version fingerprints, to verified external devices, achieving verifiable migration with consistent state; establish a secure tunnel with the external device based on the credentials, and synchronize the encrypted session context. The distributed identity and secure roaming service enables the gateway to issue verifiable credentials with time-sensitive and permission-restricted validity periods for verified user devices, establish an end-to-end encrypted tunnel based on the credentials, and synchronize the service context. The verifiable credentials include inference context digests, session key derivation parameters, and model version fingerprints, and include a fallback mechanism for cross-device migration failures and a differential synchronization mechanism for inference state snapshots under zero-trust links. The distributed identity and security roaming service also supports a fallback mechanism for cross-device migration failures, which generates a new context summary containing only the necessary state information for local recovery when roaming fails; and supports a differential synchronization mechanism for inference state snapshots, which synchronizes only the difference data between the two inference states during cross-device migrations.
[0030] In this embodiment of the invention, the signaling and data flow interaction for secure cross-device roaming is as follows: 1. The mobile app initiates a request to the command server, which includes a long-term device credential issued by the gateway TEE. 2. After verifying the credential, the command server forwards the roaming request to the REE-side service of the home gateway. 3. The REE-side service transmits the request to the distributed identity service within the TEE for authorization decision via a secure call. 4. The service within the TEE verifies the legitimacy of the request and generates a one-time session certificate valid for 5 minutes. This certificate is simultaneously sent to the command server and (through the command server) to the mobile app. The private key portion of this certificate is only generated and used within the TEE and is never exported. 5. The mobile app and the home gateway use this one-time certificate to directly establish a point-to-point encrypted tunnel via the WireGuard protocol. 6. The gateway TEE synchronizes the encrypted context of the current service session (encrypted via the tunnel key) to the mobile app through the established WireGuard tunnel, achieving seamless continuity. If synchronization fails, a fallback mechanism is initiated to generate a new context digest for local recovery. 7. The QoS guarantee mechanism dynamically adjusts transmission parameters based on network conditions to ensure low latency and reliability. 8. Differential synchronization is used, transmitting only the difference between two consecutive inference states, reducing data volume.
[0031] In this embodiment of the invention, the system further includes a Quality of Service (QoS) guarantee mechanism, used to allocate computing resources, dynamically adjust data transmission strategies, or perform task scheduling for tasks of different priorities in the Trusted Execution Environment (TEE) inference, cross-device roaming, and privacy computing service scenarios, so as to guarantee bandwidth, latency, and reliability.
[0032] In this embodiment of the invention, the privacy computing service gateway schedules external encryption tasks to be processed within the trusted execution environment, ensuring that the user's original data and model parameters do not leave the trusted execution environment, and performs QoS scheduling based on task priority and system load.
[0033] like Figure 5 As shown in the embodiment of the present invention, the internal processing flow of the external call to the privacy computing service is as follows: 1. Request phase: The external application encrypts the input data using the public key published by the gateway, attaches a task description, and sends it to the API endpoint of the privacy computing service gateway. 2. Scheduling and verification (REE side): The gateway service on the REE side verifies the API call permissions and user quota, and then forwards the encrypted task data to the TEE. 3. Secure computation (TEE side): The service within the TEE decrypts the task data, loads the user-specified model from the digital identity vault according to the description, performs the computation, and encrypts the result using the temporary public key (or session key) provided by the application. 4. Response phase: The encrypted result is returned to the gateway service on the REE side, which encapsulates it into a standard API response and returns it to the external application. The external application, the REE side system, and even the command server all access encrypted data throughout the entire process, completely preventing data leakage. 5. QoS guarantee mechanism: The privacy computing service gateway schedules tasks according to priority and system load, ensuring that high-priority tasks are processed first and meeting latency requirements.
[0034] This invention provides an AI digital identity hosting gateway system centered on "model vault + full-stack trusted inference + verifiable session credentials + private domain computing gateway," achieving: 1. Secure hosting of the entire model lifecycle: Model weights, intermediate tensors, and inference caches all run within a hardware-isolated TEE throughout their entire lifecycle, realizing a paradigm shift from "data security" to "inference security." 2. Trusted migration of AI state across devices: Secure migration of inference context is achieved through temporary session credentials derived from the TEE, ensuring the continuity and consistency of AI dialogue state across different devices. 3. Lightweight privacy computing implementation: Trusted AI capabilities are provided to third parties without exposing raw data and models, enabling low-cost privacy computing on consumer-grade hardware. 4. Multi-layered security enhancement mechanisms: These include fault recovery and replay attack protection, model version control and non-rollback, Quality of Service (QoS) guarantee mechanisms, sensitive API whitelisting mechanisms, cross-device migration failure fallback mechanisms, and differential synchronization of inference state snapshots under zero-trust links.
[0035] This invention also provides an AI digital identity hosting method based on a trusted execution environment, the method comprising the following steps: The first step is to perform security initialization and identity anchoring: complete the secure startup verification in the trusted execution environment of the gateway, generate a unique cryptographic identity for the device, and guide the user to register biometric features and encrypt and store them in the initial personal AI model; The second step is to execute local privacy-prioritized AI services: preprocess and encrypt the user input in the rich execution environment, and pass the encrypted data into the trusted execution environment; decrypt the data in the trusted execution environment and use the personal AI model to perform inference, and return the encrypted result to the rich execution environment for output; The third step is to perform cross-device secure service roaming: In response to a service request from an external terminal, the request is verified and a one-time access credential is issued within the trusted execution environment; a secure tunnel is established with the external terminal based on the credential, and an encrypted AI service context is synchronized to achieve service continuity; The fourth step is to enable external privacy capabilities: receive external encryption tasks through the privacy computing service gateway; process the task using an authorized personal AI model within the trusted execution environment; and return the encryption result to the requester.
[0036] In this embodiment of the invention, the AI service that prioritizes local privacy further includes: if the inference process is interrupted, a fault recovery mechanism is enabled to recalculate the context digest and verify the continuity of the hash chain to prevent replay attacks.
[0037] In this embodiment of the invention, the cross-device secure service roaming further includes: dynamically adjusting the quality of service policy according to network conditions; if the cross-device migration fails, enabling the fallback mechanism to generate a new context digest; and using differential synchronization to transmit only state difference data when synchronizing service contexts.
[0038] Please see Figure 1 In this specific application embodiment of the invention, a smart gateway development board using the Rockchip RK3588 chip is used as the hardware platform. This chip incorporates ARM TrustZone technology and is equipped with 4GB of LPDDR4 memory and 32GB of eMMC storage. In practical applications, large AI models (such as 7B parameter models) can be compressed to approximately 500MB using quantization technology to accommodate the gateway's storage and computing capabilities while maintaining over 90% of the original accuracy. The hardware platform of this system is based on the RK3588SoC, which internally divides into a secure world and a normal world using ARM TrustZone technology. The secure world runs an OP-TEE trusted operating system responsible for all sensitive operations; the normal world runs a custom routing system based on OpenWrt, responsible for general network tasks. The two interact through strictly defined shared memory and secure monitoring calls.
[0039] The above description only details the preferred embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0040] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.
[0041] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0042] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.
Claims
1. An AI digital identity hosting gateway system based on a trusted execution environment, characterized in that, include: The hardware platform, including the system-on-a-chip (SoC) within it, supports a hardware-level trusted execution environment. The secure firmware system runs on the hardware platform and is divided into a logically isolated trusted execution environment and a rich execution environment. The core service suite includes a digital identity vault service deployed within the trusted execution environment for securely storing users’ personal AI models and biometrics; The local secure AI inference service includes a secure inference engine running within the trusted execution environment, used to load and execute inference computations of the personal AI model in an isolated environment; Distributed identity and secure roaming service is used to generate device identities based on the trusted execution environment, issue verifiable credentials containing inference context information, and coordinate the continuation of AI services across devices; A privacy computing service gateway, deployed in the rich execution environment, is used to receive external encrypted computing tasks and schedule them to the secure inference engine within the trusted execution environment for processing. Management applications run on user terminals and are used for system configuration and interaction.
2. The AI digital identity hosting gateway system based on a trusted execution environment according to claim 1, characterized in that, The digital identity vault service also includes: A dynamic multi-factor access controller is used to enforce policy checks on access requests based on real-time biometric liveness verification. Model version control and non-rollback mechanism are used to bind the stored model version information to the digital certificate to ensure that the model version can only be upgraded in one direction; A sensitive API whitelist mechanism is used to allow only predefined and signed APIs to access sensitive resources within the vault.
3. The AI digital identity hosting gateway system based on a trusted execution environment according to claim 1, characterized in that, The local secure AI inference service also includes a model management and scheduler running in the rich execution environment; wherein the secure inference engine is configured to: receive encrypted input data from the rich execution environment, decrypt it in the trusted execution environment, use the decrypted personal AI model for inference, and return the output result after encryption; the model parameters and intermediate calculation results during the inference process do not leave the trusted execution environment at all.
4. The AI digital identity hosting gateway system based on a trusted execution environment according to claim 3, characterized in that, The local secure AI inference service also includes a fault recovery and replay attack protection mechanism, which records the step hash chain during the inference process and verifies the continuity of the hash chain during interruption recovery to prevent replay attacks.
5. The AI digital identity hosting gateway system based on a trusted execution environment according to claim 1, characterized in that, The distributed identity and secure roaming service is configured as follows: A decentralized identifier for the device is generated based on hardware information within the trusted execution environment; Issue verifiable credentials for verified external devices that have a short validity period and include an inference context digest, session key derivation parameters, and model version fingerprint. A secure tunnel is established between the credentials and the external device, and an encrypted session context is synchronized.
6. The AI digital identity hosting gateway system based on a trusted execution environment according to claim 5, characterized in that, The distributed identity and secure roaming service also includes: A cross-device migration failure fallback mechanism is used to generate a new context summary containing only the necessary state information for local recovery when roaming fails; The inference state snapshot differential synchronization mechanism is used to synchronize only the difference data between two consecutive inference states when migrating across devices.
7. The AI digital identity hosting gateway system based on a trusted execution environment according to claim 1, characterized in that, The system also includes a quality of service (QoS) guarantee mechanism, which is used to allocate computing resources to tasks of different priorities, dynamically adjust data transmission strategies, or perform task scheduling in scenarios such as trusted execution environment inference, cross-device roaming, and privacy computing services, so as to ensure bandwidth, latency, and reliability.
8. An AI digital identity hosting method based on a trusted execution environment, applied to the AI digital identity hosting gateway system based on a trusted execution environment as described in claim 1, characterized in that, The method includes the following steps: Perform secure initialization and identity anchoring: complete secure startup verification in the trusted execution environment of the gateway, generate a unique cryptographic identity for the device, and guide the user to register biometric features and encrypt and store them in the initial personal AI model; Perform local privacy-first AI services: preprocess and encrypt user input in the rich execution environment, and pass the encrypted data into the trusted execution environment; decrypt the data in the trusted execution environment and use the personal AI model for inference, and return the encrypted result to the rich execution environment for output; Perform cross-device secure service roaming: In response to service requests from external terminals, verify the requests and issue one-time access credentials within the trusted execution environment; establish a secure tunnel with the external terminal based on the credentials, and synchronize encrypted AI service contexts to achieve service continuity; Open privacy capabilities: Receive external encryption tasks through the privacy computing service gateway; process the task using an authorized personal AI model within the trusted execution environment; and return the encryption result to the requester.
9. The AI digital identity hosting method based on a trusted execution environment according to claim 8, characterized in that, The AI service that prioritizes local privacy also includes: if the inference process is interrupted, a fault recovery mechanism is enabled to recalculate the context digest and verify the continuity of the hash chain to prevent replay attacks.
10. The AI digital identity hosting method based on a trusted execution environment according to claim 8, characterized in that, The cross-device secure service roaming also includes: dynamically adjusting the quality of service policy based on network conditions; if cross-device migration fails, enabling a fallback mechanism to generate a new context summary; and using differential synchronization to transmit only state difference data when synchronizing service contexts.