Block chain credible evidence storage and AI auditing system

Abstract

The invention relates to the technical field of block chains, and discloses a block chain credible evidence storage and AI auditing system, which comprises a credible hardware module used for receiving a data hash value and a dynamic consensus variable obtained from a block chain network, and generating a dynamic data fingerprint; the terminal equipment is used for acquiring the dynamic consensus variable from the block chain network and sending the dynamic consensus variable to the trusted hardware module; the block chain platform is used for recording the transaction broadcasted by the terminal equipment; the AI auditing module is used for obtaining the recorded transaction data from the block chain platform for analysis and generating an auditing conclusion; and updating the trusted state corresponding to the trusted hardware module. According to the method, data hash and dynamic consensus variables are combined and signed through the trusted hardware module, and the data and the block chain state are subjected to cryptographic binding by utilizing the global uniqueness of the dynamic consensus variables, so that the dependence on a local clock is eliminated, and the time credibility of data storage is improved.

Description

Technical Field

[0001] This invention relates to the field of blockchain technology, specifically to a blockchain-based trusted evidence storage and AI auditing system. Background Technology

[0002] Blockchain technology, due to its decentralized, transparent, and tamper-proof characteristics, has been widely used in fields such as data storage, copyright protection, and product traceability. A common method for blockchain-based data storage involves hashing the original data to be stored, generating a unique hash value, and recording this hash value as transaction data on a distributed ledger. By comparing the hash value on the chain with the hash value of the original data, it is possible to verify whether the data has been tampered with since it was uploaded to the chain, thus ensuring its integrity to a certain extent.

[0003] However, the aforementioned traditional methods of evidence preservation still face many challenges in practical applications, mainly in the following aspects: First, there is the vulnerability of the time-anchored dimension. Traditional evidence preservation methods rely on the local system time of the data acquisition device to generate timestamps. Attackers can tamper with the device clock to perform time fraud or replay attacks on data signatures. The validity of its time proof is limited to the device itself and lacks a strong correlation with the global public trust state, resulting in insufficient credibility of the preserved data in the time dimension.

[0004] Secondly, there is the isolation of data auditing dimensions. Most existing solutions perform isolated and static verification of on-chain evidence data, that is, they only verify the integrity of the data hash and the validity of the signature. This approach cannot detect more advanced forgery behaviors. For example, attackers may use legitimate devices to submit data that is logically inconsistent or violates common sense in the domain (such as step counting). Because the single data source is formally legitimate, traditional auditing methods are difficult to effectively identify.

[0005] Third, the static nature of trust management. The trust status of data collection sources (such as hardware devices) is usually static or manually managed. Once a device is confirmed as trustworthy, all data generated by it will be trusted in form. The system lacks a dynamic, automated feedback mechanism, cannot dynamically adjust its trust level based on the device's historical data performance, and cannot automatically increase the evidence requirements when the device's trustworthiness decreases, posing a security risk. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a blockchain-based trusted evidence storage and AI auditing system, which solves the problems of low reliability of time-anchored evidence storage data, insufficient auditing capabilities against advanced forgery behaviors, and lack of dynamic feedback and adaptive adjustment capabilities in existing evidence storage schemes.

[0007] To achieve the above objectives, this invention provides a blockchain-based trusted evidence storage and AI auditing system; the system includes: Trusted hardware modules, terminal devices, blockchain platforms, and AI auditing modules; The trusted hardware module has a unique identifier and an internal private key, which is used to receive data hash values ​​and dynamic consensus variables obtained from the blockchain network, and to sign the information containing the data hash values ​​and the dynamic consensus variables using its internal private key to generate a dynamic data fingerprint.

[0008] The terminal device is connected to the trusted hardware module and is used to obtain the dynamic consensus variable from the blockchain network and send it to the trusted hardware module; receive the dynamic data fingerprint from the trusted hardware module; query the trust management smart contract deployed on the blockchain platform to obtain the current trusted state bound to the trusted hardware module; and construct a transaction containing the dynamic data fingerprint and the unique identity identifier based on the trusted state, and broadcast it to the blockchain platform.

[0009] The blockchain platform is used to record transactions broadcast by the terminal device and is equipped with the trust management smart contract, which is used to store and update the trusted state corresponding to each unique identifier.

[0010] The AI ​​audit module is used to obtain and analyze the recorded transaction data from the blockchain platform to generate audit conclusions; and acts as an oracle to call the trust management smart contract to update the trust status of the trusted hardware module according to the audit conclusions.

[0011] In a core technical solution of this invention, to address the vulnerability of time anchoring, a method is proposed to dynamically bind data collection behavior to the global consensus state of the blockchain. Specifically, when generating a signature, the trusted hardware module not only considers the data itself but also incorporates dynamic consensus variables obtained from the blockchain network (e.g., the hash value of the latest confirmed block). The specific implementation process is as follows: Hash value of the received data With dynamic consensus variables The core message to be signed is generated by concatenating the data in a predetermined order. in, This operation concatenates two byte streams in a predetermined order. This operation links the data content itself to a globally recognized time reference at the time of data acquisition.

[0012] Invoke the internal encryption unit and use its internal private key. Treating signature core messages Perform digital signature and generate dynamic data fingerprint: ;in, This represents a standard digital signature function. Due to dynamic consensus variables... The generated dynamic data fingerprint is globally unique and cannot be forged within a specific time period. It not only proved the integrity and origin of the data, but also cryptographically proved that the data must exist. Before the time point when the identified block is confirmed, attacks that forge past data evidence by tampering with local time are fundamentally prevented.

[0013] In another core technical solution of this invention, to address the problem of isolated data auditing, the AI ​​auditing module employs a multi-source data collaborative verification method based on domain knowledge. It does not treat data from different trusted hardware modules as isolated records. Instead, it first associates and aggregates evidence data from multiple different trusted hardware modules belonging to the same user based on user identity and other information, forming a multi-dimensional behavioral network. Subsequently, based on a pre-set domain knowledge graph containing causal relationships and logical rules specific to the domain, it performs cross-validation of the time and logical consistency of the multi-source heterogeneous data in the multi-dimensional behavioral network to identify logical contradictions that cannot be detected by a single data source, ultimately generating an audit conclusion.

[0014] In another core technical solution of this invention, to address the static nature of trust management, a dynamic feedback loop is constructed, from AI auditing to terminal device behavior adjustment. Specifically, before each notarization transaction is constructed, the terminal device actively queries the on-chain trust management smart contract to obtain the current trust status of its bound trusted hardware module. When the queried trust status is a preset low trust level (e.g., pending observation or suspicious), the terminal device automatically activates an enhanced notarization mode. In this mode, the terminal device forcibly collects additional enhanced supporting data (e.g., environmental images captured by its built-in camera, geographical location information obtained from the GPS, operator authentication results obtained from the biometric module, etc.), performs hash operations on the enhanced supporting data to generate an enhanced supporting data hash, and includes this hash in the transaction and broadcasts it to the blockchain. This allows the AI ​​audit module's analysis results of on-chain data to directly influence the behavior patterns of the data source, forming an adaptive security closed loop from audit risk discovery to on-chain status updates to increased terminal notarization requirements, thereby dynamically improving the security and trustworthiness of the entire system.

[0015] This invention provides a blockchain-based trusted evidence storage and AI auditing system. It has the following beneficial effects: 1. This invention combines the data hash value with the dynamic consensus variable obtained from the blockchain network through a trusted hardware module to generate a dynamic data fingerprint. Since the dynamic consensus variable is unique, publicly verifiable and tamper-proof in the global scope, the existence proof of data can be strongly bound to the evolution state of the entire distributed ledger at the cryptographic level. Its time proof validity no longer depends on the easily tampered local device clock, thereby fundamentally improving the credibility of data storage in the time dimension.

[0016] 2. This invention connects and aggregates evidence data from multiple trusted hardware modules from the same user to form a multi-dimensional behavioral network. Through a pre-built domain knowledge graph, the AI ​​auditing module can perform cross-validation of the temporal and logical consistency of multi-source data in the network, thereby identifying advanced forgery behaviors that are legitimate from a single data source but logically inconsistent when combined, thus enhancing the accuracy of the audit.

[0017] 3. This invention achieves a feedback loop through the collaborative work between the AI ​​audit module, the trust management smart contract, and the terminal device. The analysis conclusions of the AI ​​audit module can update the trust status of the hardware module recorded on the chain in real time. Before storing evidence, the terminal device will query this status and dynamically adjust the evidence storage mode according to the changes in the trust level. When the trust level decreases, it will switch to the enhanced evidence storage mode and force the collection of additional supporting data, thereby realizing dynamic response and mitigation of risks and improving the system's protection capabilities. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the system architecture of the present invention; Figure 2 This is a schematic diagram of the internal structure of the trusted hardware module of the present invention; Figure 3 This is a structural diagram of the software modules used in the terminal device of the present invention; Figure 4 This is a schematic diagram of the internal structure of the AI ​​audit module of the present invention; Figure 5 This is a flowchart of the terminal device adaptive evidence storage mode adjustment method of the present invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To better understand the present invention, the above content will be described in detail below with reference to specific embodiments.

[0021] Please see the appendix Figure 1 - Appendix Figure 5 This invention provides a blockchain-based trusted evidence storage and AI auditing system.

[0022] In one specific embodiment, the trusted hardware module is a physically independent, compact hardware device. Its external form can be adapted to different application scenarios. For example, it can be designed as a dongle that connects to external devices via a Universal Serial Bus (USB) interface, or as a box-shaped structure that directly adapts to the data port of a testing instrument via a connector. It can also be designed as an embedded module that can be directly soldered onto the main circuit board of an external device using surface mount technology (SMT).

[0023] To ensure the physical integrity and non-replicability of this trusted hardware module, its internal circuitry needs to be encapsulated in a shell with physical security mechanisms. One implementation involves using a one-piece injection molding or ultrasonic welding process for the shell, making any non-destructive disassembly impossible. In another embodiment, critical circuit units within the module are encapsulated using epoxy resin or other potting compounds. Once cured, any physical probe or modification of the internal circuitry will result in permanent damage. Furthermore, this physical security mechanism can include proactive tamper-proof measures, such as an active metal serpentine mesh or tamper-proof coating surrounding the internal circuitry. When destructive contact with this mesh or coating is detected, the module's internal logic immediately triggers an erase event, destroying critical sensitive data stored internally, such as encryption keys.

[0024] The trusted hardware module interacts with external devices through its configured communication interface. This communication interface can be a combination of one or more standardized interfaces, including but not limited to Universal Serial Bus (USB) interfaces, Universal Asynchronous Receiver / Transmitter (UART) interfaces, Serial Peripheral Interface (SPI), Internal Integrated Circuit Bus (I2C) interfaces, or General Purpose Programmable Input / Output (GPIO) pins. In this invention, these interfaces are collectively referred to as data exchange interfaces, and the specific type chosen depends on the interface standard of the external data acquisition device to be adapted, in order to achieve reliable transmission of raw data or instructions between the trusted hardware module and the external device.

[0025] The core of this module is a microcontroller (MCU), which serves as its central processing unit. This secure microcontroller preferably integrates hardware-level secure zone technology, such as ARM-TrustZone, to create a secure space and a normal space within the processor. All core firmware related to key management, cryptographic operations, and security policy execution runs in the secure space, hardware-isolated from the potentially vulnerable application logic in the normal space, thus constructing a Trusted Execution Environment (TEE).

[0026] To ensure efficient and secure cryptographic operations, this trusted hardware module can also integrate an encryption processing unit. This unit can integrate algorithms such as the Advanced Encryption Standard (AES) for symmetric encryption, secure hash algorithms (like SHA-256) for generating data digests, and elliptic curve digital signature algorithms (ECDSA) for digital signatures. Performing cryptographic calculations through this unit not only improves processing speed but also avoids the side-channel attack risks that might be encountered when implementing cryptographic operations in software on a general-purpose processor.

[0027] In one embodiment, the trusted hardware module may also integrate a secure storage unit to persistently store its identity and encryption keys. This secure storage unit can be specifically implemented as a combination of one-time programmable (OTP) memory and internal secure flash memory with access control capabilities. The globally unique identity identifier... It is permanently stored in the OTP memory at the factory, ensuring its immutability. The asymmetric private key used for signing. The private key is stored in internal secure flash memory, whose read and write access permissions are strictly controlled by the internal bus logic of the secure microcontroller. Only specific authorized code running in a trusted execution environment is allowed to access it, thus preventing any form of reading or theft of the private key from outside the module or the ordinary world.

[0028] In one specific embodiment, the trusted hardware module can be embodied as a compact hardware encryption adapter; it can be directly installed on the physical data interface of the testing instrument, and its function is to capture and process the testing data flowing through the interface in real time, generate data fingerprints and encrypt them to ensure the immutability of the original data.

[0029] Based on the aforementioned hardware configuration with physical security features, the Trusted Hardware Module (THM) can execute the core steps of the data fingerprint generation method within its internal trusted execution environment, thereby assigning initial trusted credentials to the flowing data.

[0030] One specific implementation involves the trusted hardware module first receiving data to be processed from its connected external device. In one embodiment, the external device will provide the complete raw data. The data is sent to the trusted hardware module via a data exchange interface. Subsequently, the secure microcontroller inside the trusted hardware module, within its trusted execution environment, invokes the hardware encryption unit to process the received data. Calculate its hash value. This hash value is denoted as... It is generated using the following formula: ; in, This represents a standard cryptographic hash function, such as the secure hash algorithm SHA-256. Because the hash operation is performed within the security boundary of the module, this method provides the highest level of protection for the integrity of the original data.

[0031] In another alternative implementation, to reduce the computational load and interface transmission bandwidth of the trusted hardware module, the external device or its connected terminal may pre-calculate the raw data. hash value Then only the hash value The hash value is sent to the trusted hardware module. Upon receiving the hash value, the trusted hardware module directly uses it as input for subsequent operations. (The process involves obtaining the data hash value.) Subsequently, the secure microcontroller of the trusted hardware module performs a digital signature operation to generate the underlying data fingerprint. This process specifically includes the secure microcontroller retrieving its unique, non-exportable asymmetric private key from its internal secure storage unit. Then, it invokes the encryption unit, using the private key. Data hash value Perform the signature. The underlying data fingerprint generated by this operation is denoted as... Its generation process can be expressed by the following formula: ; in, Representing standard digital signature functions, such as the Elliptic Curve Digital Signature Algorithm (ECDSA).

[0032] In one embodiment, the underlying data fingerprint This constitutes a static and reliable anchoring of the data's state at a specific collection point. It should be understood that the above two acquisition methods... Both methods (i.e., calculations performed internally by THM or pre-calculated by external devices) are valid implementations of this invention, and the choice can be adapted according to the computing power of THM, interface bandwidth, and security requirements of specific application scenarios.

[0033] After the signature operation is completed, the generated basic data fingerprint As a data packet, it is transmitted back from the trusted hardware module to the external device or terminal via the data exchange interface. This basic data fingerprint This constitutes a static and reliable anchoring of the data at a specific collection point.

[0034] Therefore, by generating this basic data fingerprint, this method can guarantee data integrity, because any infringement on the original data... Even the slightest alteration will affect its hash value. The underlying data fingerprint has changed. It failed subsequent verification. Furthermore, it ensures the authenticity of the data source because only the holder of the unique private key... Only specific trusted hardware modules can generate valid signatures, thus proving that the data has indeed been processed by this trusted source.

[0035] To address the issue that the aforementioned basic data fingerprints can be forged or replayed over time, this invention further provides a method for dynamically binding data collection behavior with the global consensus state of the blockchain. Specifically, this method elevates static point-in-time proofs to dynamic cryptographic proofs that are associated with the entire evolution of the distributed ledger.

[0036] In one specific embodiment, the trusted hardware module first obtains dynamic consensus variables from the blockchain network through its connected terminal device. Specifically, before the trusted hardware module prepares to generate a data fingerprint, the application on the connected terminal device sends a request to one or more nodes in the blockchain network to obtain the hash value of the latest confirmed block in the current network. This hash value is generated by the consensus mechanism of the entire network, is unique, publicly verifiable, and immutable within a specific time period, and therefore constitutes the dynamic consensus variable of this invention, denoted as […]. To ensure the reliability of this variable, the terminal device can request the hash value from multiple different blockchain nodes and perform a consistency comparison on the returned results. Only when multiple nodes return the same hash value is it considered a valid dynamic consensus variable and sent to the trusted hardware module through the data exchange interface.

[0037] Upon receiving valid dynamic consensus variables submitted by the terminal device Subsequently, the secure microcontroller inside the trusted hardware module will execute an enhanced signature data construction process within its trusted execution environment. This process is not limited to the data hash value. Instead of signing, the hash value of the data is compared with the received dynamic consensus variable. A deterministic combination is performed. In one specific implementation, this combination operation is the concatenation of byte streams. The constructed core message to be signed is denoted as... Its generation process is as follows: ; In the formula, represents the operation of splicing two byte streams in a predetermined order. This operation binds the content information of the original data together with the global state information of the blockchain at the time of data collection, and then forms the unique input for subsequent signature operations.

[0038] Subsequently, the secure microcontroller of the trusted hardware module retrieves the private key from its internal secure storage unit. It then invokes the encryption unit to process the previously constructed core message to be signed. A digital signature is performed. The signature generated in this process is the dynamic data fingerprint of this invention, denoted as... Its generation formula is as follows: ; in, This represents a standard digital signature function, such as the Elliptic Curve Digital Signature Algorithm (ECDSA); therefore, by using dynamic consensus variables... By incorporating signature-based core messages, this invention enables an inseparable cryptographic binding between each data collection action and the state of the blockchain ledger. The generated dynamic data fingerprint... This not only proves the integrity and origin of the data, but also demonstrates in a non-forgeable way that the original data necessarily existed in the possession of [the data source]. The identification is based on the time point before the confirmed block. Therefore, it prevents attackers from forging data signatures with past timestamps after gaining control of the device by rolling back the system clock, because attackers cannot reverse and forge a block hash that has been agreed upon by the entire network in the past, thus fundamentally improving the credibility of data fingerprints in the time dimension.

[0039] Based on the source trust capability provided by the aforementioned Trusted Hardware Module (THM), the terminal device serves as the core interface for users to interact with the system. Specifically, the terminal device is typically a portable smart device with computing and network connectivity capabilities, such as a smartphone, tablet, or dedicated handheld terminal. In this invention, the terminal device runs a core application (APP), which is functionally divided into multiple collaborative software modules.

[0040] Specifically, the terminal app may include a data acquisition module, used to establish connections with various external data acquisition devices and acquire raw data through multiple communication methods. For example, in food safety testing scenarios, this module can communicate with testing instruments equipped with THM (Thunderbolt Machine) via Bluetooth, Wi-Fi, or USB interface protocols, periodically or upon triggering specific events (such as the completion of testing) to capture test result data. For personal health management scenarios, this module can utilize Bluetooth Low Energy (BLE) to synchronize data in real-time or at set intervals with multiple peripherals such as smartwatches, smart scales, and sleep monitors, acquiring raw health data such as the user's steps, heart rate, weight, and sleep duration. .

[0041] Furthermore, the terminal app includes a device communication module dedicated to encrypted communication with the trusted hardware module (THM) for secure data and command transmission. This communication can be achieved via a USB-OTG (On-The-Go) interface, a Bluetooth secure pairing channel, or a UART / SPI interface using a specific custom protocol, for receiving raw data from external devices. or its hash value Securely send to THM and receive dynamic data fingerprint returned by THM. To prevent man-in-the-middle attacks, the device's communication module can perform two-way authentication and session key negotiation processes after establishing a connection with THM, ensuring the confidentiality and integrity of the communication content.

[0042] Furthermore, the terminal app also includes a blockchain interaction module for communicating with the blockchain network. Its main functions include: requesting the latest block hash value from blockchain nodes, constructing transactions conforming to the blockchain transaction format, signing them, and broadcasting them to the blockchain network. This blockchain interaction module can also integrate blockchain client or light client SDKs to handle functions such as blockchain node addressing, connection management, transaction encoding, wallet key management, and transaction signing. The user's private key is used for identity authentication and transaction signing on the blockchain. They are typically stored in the secure storage area of ​​the module, such as in a Hardware Security Module (HSM) or Trusted Execution Environment (TEE).

[0043] To provide a user-friendly interface, the terminal app can also include a user interface module to present the system status, operation prompts, data previews, and confirmation information for final blockchain data storage. Users can use this interface to initiate data collection requests, view real-time data, manage blockchain data permissions, and perform other system settings. For example, in a food safety scenario, inspectors can initiate the testing process, view real-time data, and authorize data storage on the blockchain after confirmation; in a personal health scenario, users can view health data trends and choose whether to share the data with third-party applications or insurance institutions.

[0044] Based on the above functional modules, the terminal device implements the data acquisition and transaction construction method as follows: The terminal app first coordinates its data acquisition module to establish a connection with one or more external data acquisition devices. Specifically, before each data acquisition task begins, the app actively scans for and connects to registered external devices. Once the connection is established, the data acquisition module continuously listens for or actively requests the external devices to transmit raw data. For example, after the testing instrument completes a sample analysis, it will automatically push the results to the terminal APP; when the smartwatch periodically synchronizes the data it collects, such as heart rate and steps, the terminal APP will capture these data streams.

[0045] Obtain the raw data After that, the device communication module of the terminal APP will... (or its hash value) It is sent to THM via a secure communication channel. Before sending, the terminal APP can... Preprocessing is performed, such as data format conversion or compression according to a preset protocol, to ensure that THM can be correctly parsed.

[0046] Receive dynamic data fingerprint from THM Afterwards, the blockchain interaction module of the terminal APP will begin to build a complete blockchain transaction system. This transaction entity contains not only the dynamic data fingerprint provided by THM. It will also integrate other necessary metadata to form a complete evidence record. This metadata typically includes: a unique identifier of the THM that generated the fingerprint. The current timestamp accurately obtained by the terminal APP This typically comes from the standard time service of the terminal device; and raw data. hash value .

[0047] Furthermore, in certain situations, to enhance the richness of data evidence, the terminal app can also include other auxiliary information. For example, when the data acquisition device has location awareness capabilities, it can include precise Global Positioning System (GPS) coordinates; when operation requires human intervention, it can include a hash of the operator's identity information. This auxiliary information can serve as... Additional fields are included.

[0048] Ultimately, the transaction body The structure is defined as follows: ; In the formula, To supplement metadata, it may include device model, calibration information, operator identification, etc. Optional enhanced corroborative data hashes are typically generated in specific modes (such as enhanced evidence storage mode).

[0049] In the transaction Once assembled, the blockchain interaction module of the terminal APP will use the user's account private key. Regarding Perform digital signature to generate the final blockchain transaction. Then, the The data is broadcast to the blockchain network, where it awaits verification by miner nodes or consensus nodes and is packaged into a new block, thus completing the on-chain notarization of the data.

[0050] Before a terminal device constructs a data transaction and prepares to upload it to the blockchain, the process it executes is not static, but dynamically adjusted according to the real-time trust status of the trusted hardware module (THM) it is bound to. This adaptive adjustment mechanism constitutes one of the core innovations of this invention.

[0051] The adaptive adjustment method is implemented as follows: When a terminal application (APP) initiates a new data storage request, its blockchain interaction module no longer directly constructs the transaction and signs it. Instead, the module first sends a request to the trust management smart contract deployed on the blockchain. Send a query request. This request contains the unique identifier of the THM currently bound to the terminal device. Smart contracts Upon receiving the request, it will search its internal storage. The corresponding latest trusted state This status is then fed back to the terminal APP via the blockchain network.

[0052] The blockchain interaction module of the terminal APP received the Then, it will be passed to the application's decision logic. This decision logic is based on... The specific value automatically executes different program branches, that is, adjusts the current data storage mode, specifically, the trusted state. It can be defined into multiple levels, such as: HIGHLY_TRUSTED, ACTIVATED, UNDER_OBSERVATION, SUSPICIOUS, LOCKED, etc.

[0053] When the queried trust status When the transaction is deemed highly reliable or activated, the terminal app will enter standard evidence storage mode. In this mode, the transaction... The construction will follow the standard process, which only includes the dynamic data fingerprint generated by THM. Data hashing THM timestamp and necessary metadata At this point, the enhanced corroborating data hash in the transaction body... Fields can remain empty or at their default values ​​because the current THM is considered highly reliable.

[0054] However, when the queried trusted status When the situation is under observation or questionable, the terminal app will immediately and forcibly activate the enhanced evidence preservation mode. In this mode, to compensate for the current decrease in THM trust level, additional auxiliary information can be requested to strengthen the evidentiary strength of the evidence. The terminal app's data acquisition module will trigger auxiliary sensors connected internally or externally to the device to collect data. For example: Environmental Image Acquisition: The app utilizes the built-in camera of the terminal device to capture images or short videos of the current data acquisition site. The app then extracts key features from these images or videos and calculates their hash values.

[0055] Location information collection: Activate the Global Positioning System (GPS) module of the terminal device to obtain the current precise geographic coordinates, and can generate an encrypted location hash value by combining the timestamp.

[0056] Operator authentication: If manual operation is required, the app can call a biometric module (such as facial recognition or fingerprint recognition) to authenticate the current operator and calculate the hash value of the authentication result. .

[0057] These additional supporting data (collectively referred to as enhanced supporting datasets, which may include image files, latitude and longitude coordinate text, biometric verification result summaries, etc.) are combined into a single dataset or constructed into a Merkle tree within the terminal device. Subsequently, a hash operation is performed on the entire dataset or its Merkle tree root to generate a unique enhanced supporting data hash. .Should It was then forcibly added to the trading platform. In this way, all supporting evidence is cryptographically bound to this evidence preservation act in an inseparable manner.

[0058] At this time, the trading body The structure will contain this non-empty Fields; If the queried trust status If the transaction is locked, the terminal app will refuse to perform any data storage operations and will display a corresponding warning message to the user. In this case, even if the terminal app attempts to construct a transaction and broadcast it to the blockchain network, the transaction will be blocked by the trust management smart contract on the blockchain. Alternatively, other consensus nodes may reject the request, thereby isolating THMs that are highly untrusted or have been identified as malicious sources.

[0059] Therefore, through the aforementioned adaptive adjustment method based on trust state, this invention can realize a dynamic, self-protective closed-loop system. Furthermore, the analysis results of the AI ​​audit module on the on-chain data can directly affect the behavior patterns of the data source, forming a complete feedback loop from AI audit to on-chain state update to terminal device behavior adjustment to affecting the next data quality, thereby effectively improving the overall security.

[0060] In one embodiment, the blockchain platform can be a consortium blockchain. Its node network topology is maintained by multiple pre-selected, mutually trusted, and authorized organizations or entities. These organizations may include data regulators, third-party auditing firms, data application providers, or THM (Thinking Machine) producers. Each authorized organization runs one or more full nodes, which collectively participate in transaction verification, block generation, and consensus maintenance.

[0061] The blockchain platform employs the Practical Byzantine Fault Tolerance (PBFT) consensus mechanism, allowing multiple nodes to reach a consensus even with the presence of some malicious nodes (Byzantine nodes). Under this mechanism, each transaction undergoes a series of pre-prepare, prepare, and commit phases before being included in a block. Specifically, when a master node proposes a new block, other nodes verify the legality of the transactions and the validity of the block, reaching a majority consensus through voting. Once more than two-thirds of the correct nodes reach a consensus on a block, it is considered finalized and added to the chain.

[0062] Furthermore, the core carrier of on-chain data is the block, whose data structure can include a series of key fields to ensure the integrity, order, and traceability of the block. The structure of each block can be represented as follows: ; in: A unique identifier for a block, typically the block height; This is the hash value of the previous block, used to link the blockchains together. The timestamp of when the block was created; This serves as the identifier for the node responsible for creating the block. This is a Merkel root hash, which is constructed from the hashes of all transactions contained in the block and is used to efficiently verify the integrity of transactions. A list of transactions included in this block; For consensus proof, under the PBFT mechanism, this is usually a set of digital signatures of the block by a majority of nodes; The transactions contained in a block are the smallest units carrying the evidence-keeping data and instructions of this invention. Each transaction... The data structure is designed as follows: ; In the formula, A unique identifier for a transaction, typically a hash value of the transaction content; The account address of the transaction initiator; The address of the transaction recipient or smart contract; The value or number of tokens attached to the transaction (if any); The timestamp when the transaction was created; The core data payload of the transaction, in this invention, is mainly used to carry the THM's evidence storage data and AI audit conclusions; The initiator of the transaction uses their private key to digitally sign the transaction content to prove the authenticity and authorization of the transaction; In its specific implementation, this invention uses the dynamic data fingerprint generated by THM. THM's unique identifier Original data hash Terminal timestamp Auxiliary metadata and enhanced supporting data hashing Information such as these is encapsulated within the Payload field of the transaction structure. The audit conclusion submitted by the AI ​​audit module... and its related Information such as this is also recorded on the blockchain through the transaction's payload field. This standardized transaction and block structure ensures that all critical information is recorded on the blockchain in a transparent, verifiable, and tamper-proof manner, providing a solid data foundation for subsequent AI auditing and trust state management.

[0063] In one embodiment, on top of the blockchain platform infrastructure, to achieve identity management and trust anchoring of the Trusted Hardware Module (THM), this invention designs and deploys a dedicated identity management smart contract. This contract serves as the foundation of trust for the entire system, responsible for maintaining the public identity information of all registered THMs and providing a transparent and verifiable query interface.

[0064] Identity Management Smart Contract Internally, it maintains a mapping table or data structure to store key identity information for each registered THM. This mapping table is typically indexed by the THM's unique identifier. As the primary key, its corresponding public key is associated with and stored. And other initialization attributes. This data structure can be represented as: mapping(bytes32?struct-THMInfo)public-thmRegistry; Among them, the bytes32 type is used for storage. THMInfo is a structure that contains Fields such as OwnerAddress (owner's on-chain address) and RegistrationTimestamp (registration timestamp) are included.

[0065] Furthermore, the contract provides a registration interface, allowing authorized entities (such as THM manufacturers or system administrators) to record new THM identity information on the blockchain. Each THM has a unique identifier upon manufacturing or initial activation. and its corresponding public key It will be submitted to through a transaction signed by an authorized entity. The contract registration function. After receiving and verifying authorization, the contract will... , Information such as the initial state is written into the thmRegistry. This process ensures that each THM has a certified and unique digital identity on the blockchain.

[0066] To support verifiers in validating THM signatures and to enable terminal devices to query THM's public key for data interaction, identity management smart contracts... It provides a public query interface. Any on-chain or off-chain entity can call these query functions, and by providing THM... You can obtain its corresponding public key. This query interface ensures Its openness, transparency, and verifiability. For example, a verifier receives a dynamic data fingerprint signed by THM. At that time, the corresponding query can be obtained by calling this query interface. Then use the public key to The signature is de-signed to verify whether it was generated by a genuine THM and whether the data has been tampered with. Therefore, this query function receives... As input, return the corresponding public key. .

[0067] In addition, identity management smart contracts It can also include a state management interface for recording and updating basic attributes of the THM, such as its activation status, user or device type. These state changes also require an authorized party's signed transaction to take effect and will be permanently recorded on the blockchain.

[0068] pass In implementing smart contracts, this invention extends the physical root of trust of THM to the digital root of trust of the blockchain. The trustworthiness of any data operation based on THM can be verified through publicly recorded on-chain data. and Verification is required.

[0069] On blockchain platforms, in addition to identity management smart contracts used to manage the underlying identity of Trusted Hardware Modules (THMs), In addition, this invention also includes a trust management smart contract. This contract is the core of the entire system's dynamic trust feedback and adaptive adjustment closed loop, responsible for maintaining and updating the dynamic trust state of each THM.

[0070] Specifically, trust management smart contracts Internally, an enumeration type (Enum) or a series of constants can be defined to represent the various trusted states of the THM. This state describes the trust level that THM is currently being assessed by the system, for example: HIGHLY_TRUSTED(0): Extremely high trust level, indicating that the device has been running well for a long time without any abnormal records; ACTIVATED(1): Activated, initial trust level, indicating that the device is registered and working properly; UNDER_OBSERVATION(2): Under observation, indicating that the device has a minor anomaly or the AI ​​audit model has detected suspicious behavior, requiring enhanced monitoring; SUSPICIOUS(3): Suspicious, indicating that there are significant anomalies or logical contradictions in the device data, and immediate measures need to be taken; LOCKED(4): Locked, indicating that the device has been confirmed to have malicious behavior or serious malfunction, and its evidence storage function has been disabled; Furthermore, the contract maintains a mapping table to store the current trusted state of each THM. This mapping table uses a unique identifier for each THM. As a key, and associated with its corresponding current trusted state. The data structure can be represented as: mapping(bytes32?uint8)-public-thmTrustStates; Among them, the bytes32 type is used for storage. The uint8 type is used to store enumerated values ​​representing trusted states. When a THM is first registered, its initial state is typically set to ACTIVATED.

[0071] and, The contract can be designed with a secure external calling interface, specifically for receiving audit conclusions from the AI ​​audit module. This interface is crucial for enabling the on-chain recording of AI audit results and driving trust state updates. To ensure security, this interface can only be called by authorized entities (i.e., the AI ​​audit module itself). Specific implementation methods include: Caller authentication: The interface internally verifies whether the caller's on-chain address msg.sender matches the preset AI audit module address (i.e., the AI ​​oracle address). Only when the two match will subsequent logic be allowed to execute.

[0072] Signature verification: The AI ​​audit module submits audit conclusions. At that time, its exclusive private key will be used. The conclusion is signed. This interface receives a message containing... After the transaction, the public key of the AI ​​auditing module stored on the blockchain will be used. (or through) (The signature is verified by querying) to ensure the authenticity and completeness of the audit conclusions.

[0073] And received a legitimate audit conclusion back, The contract will automatically execute its internal pre-defined state transition function. This is done to update the trusted state of the corresponding THM. Furthermore, this function is a defined set of logical rules, based on the current state. And AI audit conclusions To calculate the new trusted state .

[0074] ; For example, the state transition rule can be designed as follows: if It is ACTIVATEDACTIVATED, and If the data is isolated and suspicious (SUSPICIOUS), then Updated to UNDER_OBSERVATION.

[0075] if It is UNDER_OBSERVATION, and If the data is again isolated, questionable, or logically contradictory (CONTRADICTORY), then... Updated to SUSPICIOUS.

[0076] if It is SUSPICIOUS, and If it is a logical contradiction, then Updated to LOCKED.

[0077] if For UNDER_OBSERVATION or SUSPICIOUS, but For high credibility, then It can be updated back to ACTIVATED or HIGHLY_TRUSTED, implementing a trust restoration mechanism.

[0078] Furthermore, each update of the trust state is recorded as a transaction on the blockchain, thereby ensuring the transparency, traceability, and immutability of trust state changes.

[0079] Furthermore, to achieve proactive auditing and dynamic feedback of on-chain data, this invention constructs a dedicated AI auditing module. Specifically, the AI ​​auditing module logically comprises several core components. Among them, the data access module serves as the system's entry point, responsible for acquiring newly added evidence-based data from the blockchain network in real time. This data access module achieves this by listening to events broadcast by blockchain nodes, such as listening to identity management smart contracts. Registration event or trust management smart contract The system can update state events. More importantly, it can also subscribe to and parse every transaction that is included in a new block in real time. When the data acquisition module detects a new transaction containing evidence-based data as defined in this invention (e.g., its Payload field conforms to a specific data structure and contains THM), and When processing data, the original transaction data or its parsed structured data is stored in a local data storage layer (such as a distributed database or data lake). This process ensures that the AI ​​audit module can promptly obtain all newly generated data that is stored by THM.

[0080] Furthermore, when the data access module adds new evidence-based data to the blockchain, it executes a series of detailed steps, including: First, by calling the blockchain node's RPC interface or using a light client SDK, it continuously queries the header information of the latest block. When a new block is confirmed, it obtains the TransactionsList, a list of all transactions contained in that block. For each transaction in the list, the data access module parses its Payload field and extracts the unique identifier of the THM. Original data hash Dynamic data fingerprint The timestamp generated by the terminal Auxiliary metadata and enhanced supporting data hashing After parsing, this structured data is passed to a subsequent AI analysis engine for processing.

[0081] Furthermore, as another key component of the system, the oracle module is responsible for processing the audit conclusions generated by the AI ​​audit module. Secure and reliable data is written back to the blockchain to drive trust management smart contracts. Update the trusted state of THM. The oracle module also possesses its own dedicated on-chain account address and private key. The private key is also stored in the system's Hardware Security Module (HSM) or Trusted Execution Environment (TEE).

[0082] Once the AI ​​analytics engine completes its audit of a set of data and arrives at a clear audit conclusion... (For example, when determining that the data stream of a certain THM is abnormal, the conclusion could be SUSPICIOUS or UNDER_OBSERVATION.) This conclusion, along with the corresponding THM, It will be sent to the oracle module. After receiving the audit results, the oracle module will perform the following steps: Constructing the audit transaction payload: incorporating audit conclusions Affected The audit timestamp and other relevant audit context information are encapsulated into a transaction payload. .For example: ; In the formula, It is the timestamp of the audit completion. It is audit-related metadata (such as audit model version, anomaly score, etc.).

[0083] The audit payload is signed: the oracle module uses its internally stored private key. right Perform digital signature, generate signature The signature attests to the authenticity and completeness of the audit findings.

[0084] ; in, This represents a digital signature function.

[0085] In constructing and broadcasting blockchain transactions, the oracle module constructs blockchain transactions. Its target address is the trust management smart contract. The address, the transaction payload contains and The data is processed and the final transaction signature is performed using the account's private key from the oracle module.

[0086] And that The transaction is broadcast to the blockchain network. Once the transaction is verified by network nodes and confirmed through the consensus mechanism, the trust management smart contract... The system will then receive this audit conclusion and update the trust status of the corresponding THM accordingly, completing the closed loop from audit to feedback.

[0087] Furthermore, in the core processing stage of the AI ​​audit module, the multi-source data collaborative verification and causal inference method of this invention can identify more complex and concealed advanced forgery behaviors.

[0088] Specifically, this method no longer treats data from different Trusted Hardware Modules (THMs) as isolated records. Instead, the AI ​​audit module's data processing module first correlates data from different THMs based on user identity information stored on or off the blockchain. For example, by querying identity management smart contracts. They will belong to the same user All THMs (such as smartwatches) and smart scale The system then identifies the evidence data generated by these THMs (including their dynamic data fingerprints). Original data hash timestamp (etc.) are aggregated and organized according to time sequence to build a multidimensional behavioral network for each user.

[0089] In a specific implementation scenario, suppose the user It has a smartwatch for recording steps (linked to THM_Watch) and a smart scale for measuring weight (linked to THM_Scale). The AI ​​auditing module continuously pulls step data generated by these two THM devices from the blockchain. and weight data This data, along with their respective dynamic data fingerprints and timestamps, is aggregated into the user's... In the multidimensional behavioral network, a composite dataset is formed with time as the axis and data type as the dimension.

[0090] Building upon this multidimensional behavioral network, the AI ​​model utilizes a pre-built domain knowledge graph to cross-validate the temporal and logical consistency among multi-source data. Furthermore, this domain knowledge graph is a structured knowledge base containing domain-specific expertise and causal relationships. For example, in the field of personal health management, the knowledge graph can encode the following causal or correlational rules in physiology or behavior: Rule 1: A significant increase in steps within a short period of time ( This is usually accompanied by an increase in heart rate.

[0091] Rule 2: If a user claims to have engaged in strenuous exercise, their weight ( ) will decrease shortly after the exercise ends. A slight decrease occurred (due to moisture loss).

[0092] Rule 3: If a user claims to exercise regularly over a long period of time, their weight curve should show a specific trend (e.g., with a controlled diet, weight should decrease slowly or remain stable).

[0093] Rule 4: An unusually high number of steps at night (such as between 2 a.m. and 6 a.m.) is inconsistent with normal human physiological activity.

[0094] Furthermore, the AI ​​auditing model can be a rule-based expert system, or a complex model that combines machine learning algorithms (such as anomaly detection models or graph neural networks). It can also call upon the rules in these knowledge graphs to perform cross-validation and causal inference on multi-source data.

[0095] Specifically, in scenarios involving identifying logically inconsistent step-counting behavior: Data collection and preprocessing: The AI ​​model extracts recent (e.g., the past 24 hours) smartwatch step count data from the user's multidimensional behavioral network. Heart rate data Specifically: Feature engineering: From and Extract time series features, such as hourly step increments. Heart rate variability .

[0096] Rule matching and causal inference: The AI ​​model applies rule 1 from the knowledge graph. It then identifies whether a time window exists. to within that window Significantly above average, but However, it remained stable, or even declined.

[0097] ; in, Represents the average value; The value is significantly greater than; This means approximately equal to.

[0098] For example, if the number of steps suddenly increases from 500 to 5000 in a certain hour, but the heart rate data recorded by the smartwatch during the same hour shows that the user is in a resting state and there is no obvious fluctuation, this constitutes a phenomenon that contradicts common sense in physiology. In this case, a single data source can only report the surge in steps, but cannot determine its rationality; only by combining heart rate data for cross-validation can the abnormality be revealed.

[0099] If the AI ​​model, through this multi-source collaborative verification and causal inference, discovers significant logical contradictions or violations of pre-defined causal relationships among the data, it will classify this as an abnormal behavior. At this point, the AI ​​auditing module will generate an audit conclusion. For example, it can be labeled as SUSPICIOUS or UNDER_OBSERVATION, and this conclusion can be linked to the relevant THM. It is passed to the oracle module, ready to be written to the blockchain.

[0100] Furthermore, by fusing multi-source heterogeneous data and leveraging domain knowledge for advanced reasoning, this method can effectively identify sophisticated forgery behaviors that are planned or attempt to evade detection through a single data source.

[0101] In one specific embodiment, after the AI ​​audit module completes the collaborative verification and causal inference of multi-source data and generates audit conclusions, it can use the AI ​​system as an oracle to provide proactive feedback, which is used to securely and reliably write these intelligent analysis results back to the blockchain to drive the operation of the entire trust loop.

[0102] Specifically, the implementation of this method begins with the AI ​​audit module generating audit conclusions. Then, the conclusion is digitally signed. Specifically, the AI ​​analysis engine will audit the conclusion. (This is typically an enumeration value, such as SUSPICIOUS or UNDER_OBSERVATION), along with the unique identifier of the audited Trusted Hardware Module (THM). timestamp of the audit and any related audit metadata Organize into a message awaiting signature The message structure can be defined as follows: ; Subsequently, the oracle module of the AI ​​auditing module will use its pre-generated and securely stored private key. Regarding this message Perform a digital signature. The result of this signature operation is denoted as... Its generation process can be represented as: ; In the formula, This represents a standard digital signature function, such as the Elliptic Curve Digital Signature Algorithm (ECDSA); and this digital signature... This ensures the authenticity of the information source for the audit conclusion, proving that the conclusion was indeed issued by the AI ​​audit module and not by any other unauthorized entity; it also guarantees the integrity of the audit conclusion, proving that the conclusion has not been tampered with in any way since it was signed. The public key of the AI ​​audit module... It will be registered on the blockchain during system deployment (e.g., through an identity management smart contract). (This information is available for public inquiry and verification by all participants.)

[0103] After the signature is completed, the oracle module of the AI ​​audit module submits the signed audit conclusion to the blockchain. This process involves the trust management smart contract on the blockchain. External calls. The specific steps are as follows: 1. Constructing a Blockchain Transaction: The oracle module constructs a new blockchain transaction. The sender of this transaction is the oracle module's own on-chain account address, OracleAddress, and the receiver is the trust management smart contract. The on-chain deployment address. The transaction's core payload will contain the aforementioned signature message. and its digital signature .

[0104] 2. Calling the smart contract interface: The oracle module calls the smart contract interface. The secure external interface pre-defined in the contract, specifically for receiving audit conclusions, will and Submitted as a parameter.

[0105] 3. Broadcasting transactions to the blockchain network: Constructing a complete transaction The transaction is broadcast through the blockchain network's client SDK or RPC interface. Miner nodes or consensus nodes in the network then verify the transaction's legitimacy (including the sender's signature, transaction fees, etc.) and package it into a new block.

[0106] once The trust management smart contract was successfully packaged and added to the blockchain. The system will then receive the transaction and automatically execute its internal trust state update logic. First, the identity of the caller (i.e., the OracleAddress of the oracle module) will be verified, and the on-chain stored... verify The validity of the contract. Once verified, the contract will proceed according to... Audit conclusions and the preset state transition function Update the corresponding Trusted state .

[0107] Thus, through this series of steps, the AI ​​audit module can achieve a closed loop from intelligent analysis to on-chain feedback, enabling AI's intelligent decisions to directly and securely influence the trust status of THM in the system, thereby triggering the adaptive adjustment behavior of terminal devices and improving the overall system's response speed and defense capabilities.

[0108] In one application example, for instance, regarding the detection results of potential prohibited veterinary drug residues (such as nitrofurans) in food, the process of uploading the detection results after the test is completed is as follows: Trusted data source anchoring: On-site inspectors use a portable rapid testing instrument physically connected to a Trusted Hardware Module (THM) to test food samples. The instrument generates raw test result data (e.g., sample number A, AOZ residue: 1.5 g / kg) and calculates its data hash value. .

[0109] Generating dynamic data fingerprints: The terminal device operated by the inspector (such as a dedicated tablet computer) obtains the hash value of the latest confirmed block from the blockchain platform in real time as a dynamic consensus variable. And send it to THM. THM will internally... and After being concatenated, it is signed using its internal private key to generate a unique dynamic data fingerprint. The fingerprint is then returned to the terminal device. This fingerprint cryptographically proves that the detection result existed before a specific block time point and originated from that specific detector.

[0110] Trust state adaptive adjustment: Before constructing an on-chain transaction, the terminal device will query the trust management smart contract on the chain to obtain the current trust state of THM.

[0111] Standard mode: If the status is highly trustworthy, the terminal device directly constructs a standard transaction.

[0112] Enhanced Evidence Preservation Mode: If the detector is flagged as needing observation by the AI ​​audit module due to anomalies in historical data, the terminal device automatically activates Enhanced Evidence Preservation Mode. In this mode, the application will require the operator to use a tablet to take photos of samples with a time watermark, record the GPS coordinates of the testing location, and may require fingerprint or facial recognition verification of the operator's identity. This enhanced corroborating data is hashed to generate an Enhanced Corroborating Data Hash. .

[0113] Transaction Construction and On-Chain: Terminal devices will use THM's unique identifier and dynamic data fingerprint. Data hashing And generated in enhanced mode The information is packaged into a transaction, signed by the inspector, and broadcast to the blockchain platform to complete the notarization.

[0114] AI-powered intelligent auditing and feedback: The AI ​​auditing module continuously acquires data from the blockchain. It not only verifies the validity of individual data points but also performs multi-source data collaborative verification. For example, it correlates the nitrofuran test result of a sample with other test data from the same batch of samples (such as chloramphenicol results from another device). If the AI ​​detects multiple reports from a particular inspector showing GPS anomalies at the testing locations, or logical inconsistencies in the test results of a batch of samples from multiple different locations, it generates a questionable audit conclusion. This conclusion securely invokes the trust management smart contract through the oracle module, automatically downgrading the trust status of the involved THM or operator account, thereby triggering stricter evidence requirements for future testing.

[0115] Through the above steps, the data collection behavior is strongly bound to the global state of the blockchain, and combined with the multi-dimensional logical verification and dynamic trust feedback loop of AI, a new stage is entered to ensure the authenticity, accuracy and immutability of the uploaded test results; In another application embodiment, besides the potential presence of prohibited veterinary drug residues (such as nitrofurans), the system described in this invention can also be used to process data when detecting food additives, pesticide residues, heavy metal contamination (such as lead and mercury), and pathogenic microorganisms (such as Salmonella) in food. Regardless of whether the data content is concentration values, compliance conclusions with limits, or microbial counts, the system treats it as raw data to be verified. A dynamic data fingerprint, bound to the blockchain state, is generated through a trusted hardware module and ultimately linked to the continuous supervision of the AI ​​audit module, thus providing a unified and high-standard credibility guarantee for food safety testing across different dimensions.

Claims

1. A blockchain trusted notarization and AI auditing system, characterized in that, The application comprises: a trusted hardware module having a unique identity identifier and an internal private key, configured to receive a data hash value and a dynamic consensus variable obtained from a blockchain network, and sign information containing the data hash value and the dynamic consensus variable using the internal private key to generate a dynamic data fingerprint; a terminal device connected to the trusted hardware module, configured to obtain the dynamic consensus variable from the blockchain network and send it to the trusted hardware module, receive the dynamic data fingerprint from the trusted hardware module, query a trust management smart contract deployed on a blockchain platform to obtain a current trusted state bound to the trusted hardware module, and construct a transaction containing the dynamic data fingerprint and the unique identity identifier based on the trusted state and broadcast it to the blockchain platform; a blockchain platform configured to record the transaction broadcast by the terminal device and deploy the trust management smart contract, wherein the trust management smart contract is configured to store and update the trusted state corresponding to each unique identity identifier; an AI audit module configured to analyze the recorded transaction data obtained from the blockchain platform to generate an audit conclusion, and call the trust management smart contract as an oracle to update the trusted state corresponding to the trusted hardware module according to the audit conclusion. The trusted hardware module comprises:

2. The blockchain trusted archiving and AI auditing system of claim 1, wherein, a secure microcontroller integrated with a trusted execution environment; a secure storage unit for persistently storing the internal private key; a hardware encryption operation unit for performing encryption operations. The trusted hardware module is specifically configured to: 3.The blockchain trusted archiving and AI auditing system of claim 2, wherein, splice the data hash value and the dynamic consensus variable in a predetermined order to generate a core message to be signed; call the encryption operation unit to use the internal private key to digitally sign the core message to be signed to generate the dynamic data fingerprint. The blockchain platform further comprises an identity management smart contract configured to store the unique identity identifier of each trusted hardware module and the public key corresponding to the unique identity identifier, and provide a public query interface.

4. The blockchain trusted archiving and AI auditing system of claim 1, wherein, The AI audit module comprises a data access module configured to:

5. The blockchain trusted archiving and AI auditing system of claim 1, wherein, listen to and parse the payload of a new transaction in the blockchain platform to extract the unique identity identifier, the data hash value and the dynamic data fingerprint. The AI audit module is further configured to:

6. The blockchain trusted archiving and AI auditing system of claim 5, wherein, based on user identity information, correlate and aggregate the evidence data of multiple different trusted hardware modules from the same user to form a multi-dimensional behavior network; based on a preconfigured domain knowledge graph, cross-verify the multi-source data in the multi-dimensional behavior network for time and logical consistency to generate the audit conclusion. The AI audit module further comprises an oracle module comprising a dedicated on-chain account address and a private key, and is configured to:

7. The blockchain trusted archiving and AI auditing system of claim 1, wherein, use the private key to digitally sign an audit message containing the audit conclusion and the corresponding unique identity identifier; construct a blockchain transaction to call the trust management smart contract to submit the signed audit message. The terminal device is further configured to:

8. The blockchain trusted archiving and AI auditing system of claim 1, wherein, ​ When it is queried that the trusted state is a preset low trust level, an enhanced evidence storage mode is activated; In the enhanced evidence storage mode, additional enhanced corroboration data is collected, and the enhanced corroboration data is subjected to a hash operation to generate enhanced corroboration data hash; And the enhanced corroboration data hash is included in the transaction for broadcast.

9. The blockchain trusted archiving and AI auditing system of claim 8, wherein, The enhanced corroboration data includes: An environmental image of a current data collection site captured by a built-in camera of the terminal device; Geographic location information obtained by a global positioning system module of the terminal device; An operator identity verification result obtained by a biometric identification module of the terminal device.

10. The blockchain trusted archiving and AI auditing system of claim 7, wherein, The trust management smart contract further includes a secure external calling interface, and the interface is used to: Verify whether a transaction caller address is an on-chain account address of the oracle module; Verify a digital signature of the oracle module on the audit message; And after verification passes, a preset state transition function is executed according to the received audit conclusion to update the trusted state corresponding to the trusted hardware module.

Images