A medical blockchain storage and AI handwriting dynamic analysis system based on a national secret SM2 signature

Abstract

The application relates to the fields of medical information security and blockchain storage, and discloses a medical blockchain storage and AI handwriting dynamic analysis system based on a national secret SM2 signature, which comprises a biological entropy generation module configured to collect handwriting dynamic sequences in a writing process, wherein the handwriting dynamic sequences at least include pen pressure, acceleration and grip time sequence data; a feature vector is extracted from the handwriting dynamic sequences, and the feature vector is converted into 256-bit uniformly distributed biological entropy Ebio through an SHA-3 chaotic expansion function. In the application, the handwriting dynamic features are used to generate biological entropy, which is mixed with a hardware true random number to serve as an SM2 signature random number source, so that the cryptographic binding of the signature and the behavior features is realized, even if the private key is leaked, the signature cannot be forged; the signature and handwriting data are prevented from being attacked through stripping through double signature verification; the handwriting drift is monitored through a Mahalanobis distance, and the signature verification threshold is adaptively adjusted, so that the false rejection rate of the legal signature is reduced.

Description

Technical Field

[0001] This invention relates to the fields of medical information security and blockchain evidence storage technology, and in particular to a medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic standard SM2 signature. Background Technology

[0002] With the deepening development of medical informatization, the authenticity and non-repudiation of medical behavioral data such as electronic prescriptions, electronic medical records, and surgical records have become key requirements for legal and clinical management. In existing technologies, digital signature technology is usually used to store medical data. Among them, the Chinese cryptographic SM2 signature algorithm has been gradually promoted and applied in the medical field due to its advantages such as independent controllability and high security. At the same time, AI handwriting dynamic analysis technology is used to verify the legitimacy of the signer's identity.

[0003] However, the existing system has the following technical shortcomings: First, SM2 signatures and handwriting analysis data are stored independently on the blockchain, and there is no cryptographic binding relationship between the two. Attackers can replace or forge dynamic handwriting data without destroying the SM2 signature, causing the evidence to become invalid and making it impossible to truly achieve the inseparability of the signer's biometrics and signing behavior. Second, the existing SM2 signature algorithm relies on hardware true random number and pseudo random number generators for random number generation. Pure hardware solutions are costly, while pure pseudo random solutions have predictability risks. Furthermore, neither of them utilizes the signer's own behavioral biometrics, limiting the potential for security improvement. Third, existing handwriting recognition systems use fixed thresholds to determine authenticity, without considering the natural handwriting drift caused by doctors' aging, fatigue, injury, or neurodegenerative diseases. The fixed thresholds lead to a gradual increase in the proportion of legitimate signatures being misjudged as forgeries over time. Fourth, the existing system's handwriting dynamic data is only used for identity verification and does not fully explore the doctor's health information contained within it. The function is limited and the data value is not fully utilized.

[0004] In summary, the existing technology lacks a medical evidence storage system that can cryptographically bind SM2 signatures to dynamic handwriting features, while also possessing adaptive threshold adjustment and health monitoring functions. Summary of the Invention

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature, comprising: The bio-entropy generation module is configured to collect dynamic sequences of handwriting during the writing process, wherein the dynamic sequences of handwriting include at least time-series data of pen pressure, acceleration, and grip force; extract feature vectors from the dynamic sequences of handwriting, and convert the feature vectors into 256-bit uniformly distributed bio-entropy Ebio using the SHA-3 chaotic expansion function; and output the bio-entropy Ebio and its hash commitment H(Ebio). The SM2 signature random number mixing module is configured to obtain a hardware true random number RTRNG; calculate the random number k of the SM2 signature algorithm = (RTRNG XOR Ebio) mod n, where n is the order of the SM2 elliptic curve; generate an SM2 signature pair (r,s) using the random number k; and upload the signature pair (r,s) and the hash commitment H(Ebio) to the blockchain for evidence storage. The signature verification module is configured to obtain the signature pair (r, s) to be verified and the hash commitment H (Ebio) from the blockchain; verify the mathematical correctness of the SM2 signature pair; and verify whether the hash commitment in this signature is consistent with the hash commitment recorded on the chain; the signature is deemed valid only if both pass.

[0006] As a further description of the above technical solution: The biological entropy generation module also includes a degradation mode. When the data quality of the handwriting dynamic sequence is lower than a preset threshold, it automatically switches to a pure hardware true random number mode, that is, sets k=RTRNG mod n, and records this degradation event on the blockchain.

[0007] As a further description of the above technical solution: The SM2 signature private key is managed using a (2,2) threshold secret sharing scheme. The first fragment of the private key is stored on the doctor's mobile terminal, and the second fragment is stored in the hardware security module. When generating a complete signature, both fragments need to participate simultaneously; if either fragment is missing, a valid signature cannot be generated.

[0008] As a further description of the above technical solution: It also includes a handwriting drift detection and adaptive threshold module, configured to construct a reference distribution using historical handwriting features within a 90-day rolling time window, including a mean vector μref and a covariance matrix Σref; Calculate the Mahalanobis distance Ddrift between the current signature handwriting feature vector xt and the reference distribution; dynamically adjust the judgment threshold of the signature verification module according to the value of the Mahalanobis distance Ddrift: when Ddrift≤1.5, use the standard threshold of 0.85; when 1.5<Ddrift≤2.5, record the drift log and keep the standard threshold unchanged; when 2.5<Ddrift≤3.5, adjust the signature verification threshold to 0.75 and trigger manual sampling review; when Ddrift>3.5, adjust the signature verification threshold to 0.70 and generate a doctor's health warning.

[0009] As a further description of the above technical solution: The method for generating the doctor's health warning is that the system uses a zero-knowledge scope proof protocol to generate a proof that Ddrift > 3.5 and the probability of neurodegenerative disease risk of handwriting feature mapping is greater than 0.6. The zero-knowledge proof is stored on the blockchain, and the doctor can use their private key to decrypt and view detailed warning information.

[0010] As a further description of the above technical solution: The proportion of manual review of the samples is 1% to 5% of all signatures. The reviewers make the judgment by comparing the pressure distribution map and the trajectory heat map. When the manual review result is inconsistent with the model judgment, the review result is updated with a weight of 1% to update the reference distribution.

[0011] As a further description of the above technical solution: The SHA-3 chaotic expansion function uses the SHAKE-256 algorithm. The input is a 128-bit original feature quantization value, and the output is a 256-bit uniformly distributed biological entropy Ebio. The Shannon entropy value of the output entropy is not less than 0.99.

[0012] As a further description of the above technical solution: The SM2 signing private key adopts a periodic rotation mechanism, automatically generating a new SM2 key pair every two years. The old private key signs the new public key and is recorded on the chain as proof of key rotation. The old and new public keys are stored in the blockchain at the same time. New transactions are signed using the new private key, while historical transactions are still verified using the old public key.

[0013] As a further description of the above technical solution: The grip force data is collected by a piezoresistive or capacitive grip force sensor integrated in the smart pen, with a sampling frequency of not less than 100Hz and a quantization accuracy of not less than 12 bits; when the grip force variance of 10 consecutive sampling points is lower than a preset threshold, the grip force signal is determined to be invalid, triggering the degradation mode described in claim 2.

[0014] The present invention has the following beneficial effects: 1. This invention generates biological entropy by extending the dynamic features of handwriting with SHA-3 chaos, and mixes it with hardware true random numbers via XOR as the random number source for SM2 signature. This achieves a cryptographic level binding between the signature and the biometric features of the signer. Even if the SM2 private key is leaked, attackers cannot forge a valid signature if they do not have the dynamic handwriting data of the legitimate signer. This helps to improve the non-repudiation capability of medical evidence.

[0015] 2. This invention adds consistency verification of bio-entropy hash commitment to the signature verification module, forming a dual verification mechanism of "mathematical correctness + biometric consistency", which effectively prevents the attack of separating signature and handwriting data and fills the security vulnerability of decoupled storage of the two in the prior art.

[0016] 3. This invention introduces a handwriting drift monitoring and adaptive threshold adjustment module, which quantifies the natural changes in doctors' handwriting through Mahalanobis distance and dynamically adjusts the signature verification threshold, reducing the false rejection rate of legitimate signatures caused by age, fatigue and disease, while retaining the ability to detect abnormal signatures.

[0017] 4. This invention makes full use of health information in dynamic handwriting data. When abnormal drift is detected, it automatically generates a health warning in the form of zero-knowledge proof and stores it on the blockchain. This provides doctors with early screening function for neurodegenerative diseases and realizes the functional upgrade from "identity verification" to "identity verification + health monitoring".

[0018] 5. This invention manages SM2 private keys through a threshold secret sharing scheme, storing private key fragments in the mobile terminal and hardware security module respectively. Signing cannot be completed if any fragment is missing, effectively preventing single point of leakage risk. At the same time, a degradation mode and a periodic rotation mechanism are designed to ensure the availability and security of the system in scenarios where sensors fail or the system is in long-term operation. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall architecture proposed in this invention; Figure 2 This is a flowchart of the SM2 signature generation and storage process proposed in this invention; Figure 3 This is a schematic diagram of the adaptive threshold adjustment process for signature verification and handwriting drift proposed in this invention. Detailed Implementation

[0020] The technical solutions of 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.

[0021] Example 1 This embodiment provides a medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature.

[0022] The system in this embodiment includes the following modules: a biological entropy generation module, an SM2 signature random number mixing module, a signature verification module, a handwriting drift monitoring and adaptive threshold module, and a key management unit.

[0023] The bio-entropy generation module is configured to collect dynamic sequences of handwriting during the writing process, and the dynamic sequences of handwriting include at least pen pressure time series data, acceleration time series data, and grip force time series data.

[0024] Specifically, the pen pressure sensor is a resistive thin-film pressure sensor with a range of 0 to 500 grams and an accuracy of ±2 grams, which is installed at the tip of the smart pen. The accelerometer is a triaxial microelectromechanical system accelerometer with a range of ±8 times the acceleration due to gravity and a sampling frequency of 200 Hz. The grip force sensor is a piezoresistive thin-film sensor integrated into the grip area of ​​the smart pen with a sampling frequency of 100 Hz and a quantization accuracy of 12 bits.

[0025] During the writing process, the aforementioned sensors continuously collect data in a time-synchronized manner, forming pen pressure sequences, triaxial acceleration sequences, and grip force sequences. The bio-entropy generation module extracts feature vectors from these time sequences. These feature vectors include velocity spectrum, acceleration spectrum, pressure entropy, and average grip force. Specifically, the extraction method is as follows: the velocity spectrum is obtained by integrating the acceleration sequence; the acceleration spectrum is obtained by performing a Fourier transform on the acceleration sequence itself; the pressure entropy is calculated based on the probability distribution of the pen pressure sequence; and the average grip force is obtained by taking the arithmetic mean of the grip force sequence.

[0026] Then, the biological entropy generation module converts the above feature vector into a 256-bit uniformly distributed biological entropy, denoted as Ebio, through the SHA-3 chaotic expansion function. In this embodiment, the SHA-3 chaotic expansion function adopts the SHAKE-256 algorithm, with a 128-bit original feature quantization value as input and a 256-bit biological entropy as output. After testing, the Shannon entropy value of the output entropy is not less than 0.99, which has good uniform distribution characteristics.

[0027] The biological entropy generation module also calculates the hash commitment of the biological entropy, denoted as H(Ebio). The hash algorithm adopts the national cryptographic SM3 algorithm, and the module finally outputs the biological entropy Ebio and its hash commitment H(Ebio) for use by subsequent modules.

[0028] The SM2 signature random number mixing module is configured to obtain hardware true random numbers, denoted as RTRNG. In this embodiment, the hardware true random number source adopts a chip-level thermal noise amplifier circuit with an output rate of 1 megabit per second, which meets the requirements of the national cryptography industry standard GM / T 0062.

[0029] The module calculates the random number k used in the SM2 signature algorithm using the following formula: k = (RTRNG XOR Ebio) mod n Where n is the order of the SM2 elliptic curve, and XOR represents the bitwise XOR operation.

[0030] The module uses the calculated random number k to generate a signature pair (r, s) according to the SM2 signature algorithm standard. The SM2 private key used in the signing process is stored and retrieved according to the key management method described later in this embodiment.

[0031] After the signature is generated, the module uploads the signature pair (r, s) and the previously generated hash commitment H (Ebio) to the blockchain for storage. In this embodiment, the blockchain adopts the Hyperledger Fabric consortium blockchain framework, the consensus mechanism is Raft, and the block generation interval is 2 seconds.

[0032] The signature verification module is configured to obtain the signature pair (r, s) to be verified and the corresponding hash commitment H (Ebio) from the blockchain. The signature verification process includes two independent verification steps: The first check is the mathematical correctness check of the SM2 signature pair. The signature verification module uses the signer's public key to perform mathematical verification on (r, s) according to the SM2 signature verification algorithm to determine whether the signature was signed by the corresponding private key.

[0033] The second verification is the hash commitment consistency verification. The signature verification module compares the hash commitment carried in this signature with the hash commitment recorded on the blockchain to determine whether the two are completely consistent.

[0034] The signature verification module determines the signature to be valid only if both of the above checks pass. If either check fails, the signature is determined to be invalid and the abnormal event is recorded.

[0035] The biological entropy generation module also includes a degradation mode, where the system monitors the data quality of the dynamic handwriting sequence in real time.

[0036] In this embodiment, the criteria for determining that the data quality is lower than the preset threshold are: the pen pressure variance of 15 consecutive sampling points is less than 0.5 g squared, or the grip strength variance of 10 consecutive sampling points is less than 0.1 N squared.

[0037] When the above conditions are met, the system determines that the quality of the handwriting dynamic data is unreliable and automatically switches to pure hardware true random number mode. In this mode, the calculation formula for the SM2 signature random number k is simplified to: k = RTRNG mod n This means that biological entropy will no longer be mixed. Simultaneously, the system will record this downgrade event as a blockchain transaction, containing a timestamp of the downgrade occurrence, the device identifier, and the downgrade reason code. This record is immutable and facilitates subsequent auditing.

[0038] The SM2 signature private key is managed using a (2,2) threshold secret sharing scheme. Specifically, the private key is split into two fragments: the first fragment is stored in the doctor's mobile terminal, which is a security-hardened smartphone, and the fragment is stored in a secure enclave of the terminal; the second fragment is stored in a hardware security module, which conforms to the national cryptographic industry standard GM / T 0045.

[0039] When a complete signature needs to be generated, the signature module must call both slices simultaneously for computation. If either slice is missing, a valid signature cannot be generated. If the mobile terminal is offline or the hardware security module is unavailable, the signature operation will fail and return an error message.

[0040] The system also includes a handwriting drift monitoring and adaptive threshold module. This module is configured to construct a reference distribution using historical handwriting features within a 90-day rolling time window. The reference distribution includes a mean vector μref and a covariance matrix Σref. Specifically, the construction method is as follows: extract the handwriting feature vectors corresponding to all valid signatures of the doctor in the past 90 days, calculate the one-dimensional mean of these vectors to form μref, and calculate the covariance matrix to form Σref.

[0041] For the current signature, the module extracts its handwriting feature vector xt and calculates the Mahalanobis distance Ddrift between xt and the reference distribution. The calculation formula is as follows: Based on the value of the Mahalanobis distance Ddrift, the module dynamically adjusts the judgment threshold of the signature verification module, and the specific rules are as follows: When Ddrift ≤ 1.5, the standard threshold of 0.85 is used without any additional processing.

[0042] When 1.5 < Ddrift ≤ 2.5, record one handwriting drift log, keeping the standard threshold of 0.85 unchanged.

[0043] When 2.5 < Ddrift ≤ 3.5, the signature verification threshold is adjusted to 0.75, and a manual review is triggered. The proportion of manual review is 3% of all signatures. The reviewers make judgments by comparing the pressure distribution map and the trajectory heat map. When the manual review results are inconsistent with the automatic judgment of the model, the review results are updated with a weight of 1% to update the reference distribution.

[0044] When Ddrift > 3.5, the signature verification threshold is adjusted to 0.70, and a doctor's health warning is generated.

[0045] The doctor's health warning is generated as follows: The system first uses a trained neurodegenerative disease risk model to map handwriting features to risk probabilities. This model is based on a multilayer perceptron network, with 128-dimensional handwriting features as input and risk probabilities between 0 and 1 as output. When Ddrift > 3.5 and the risk probability output by the model is greater than 0.6, the system determines that a warning needs to be issued.

[0046] To protect physician privacy, the system uses a zero-knowledge scope proof protocol to generate proofs. This embodiment employs the Bulletproofs protocol, and the generated proof demonstrates that both of the following conditions are met simultaneously: the Mahalanobis distance (Ddrift) is greater than 3.5, and the probability of neurodegenerative disease risk is greater than 0.6. This proof does not disclose the specific Ddrift value or the probability of disease risk. The proof generation time is no more than 1.5 seconds, and the proof size is no more than 1KB.

[0047] The system stores the zero-knowledge proof on the blockchain and simultaneously encrypts the detailed warning information (including specific Ddrift values, risk probabilities, time series charts, etc.) using the doctor's personal public key and stores it in an off-chain or on-chain encrypted field. The doctor can use their own private key to decrypt and view the detailed warning information. Any other party can only verify the validity of the zero-knowledge proof but cannot obtain the specific health data.

[0048] The SM2 signing private key employs a periodic rotation mechanism; every two years, the system automatically generates a new SM2 key pair. The old private key is used to digitally sign the new public key, and this signature serves as proof of the key rotation, stored on the blockchain. Both the old and new public keys are simultaneously stored in the blockchain, forming a public key chain. Thereafter, newly generated transactions are signed using the new private key, while historical transactions are still verified using the old public key. When verifying transactions, the signature verification module automatically selects the corresponding public key based on the transaction's timestamp.

[0049] Grip force data is collected via a piezoresistive or capacitive grip force sensor integrated into the smart pen. In this embodiment, a piezoresistive sensor is used, with a sampling frequency of no less than 100 Hz and a quantization accuracy of no less than 12 bits. The system monitors the validity of the grip force signal in real time: when the variance of the grip force at 10 consecutive sampling points is lower than a preset threshold (e.g., 0.05 N²), the grip force signal is determined to be invalid, and the aforementioned degradation mode is triggered.

[0050] Example 2 The differences between this embodiment and Embodiment 1 are as follows.

[0051] First, the parameters of the biological entropy generation module were adjusted. The sampling frequency of the grip force sensor was adjusted to 150 Hz, the quantization accuracy was 10 bits, the input of the SHAKE-256 algorithm was 160 bits of original feature quantization value, and the output was still 256 bits of biological entropy. The Shannon entropy value of the output entropy reached 0.995.

[0052] Second, the triggering conditions for the downgrade mode have been adjusted. When the pressure variance of 15 consecutive sampling points is less than 0.5 g squared, or the grip force variance of 10 consecutive sampling points is less than 0.1 N squared, it will automatically switch to pure hardware true random number mode.

[0053] Third, the proportion of manual review in the adaptive threshold module is adjusted to one percent of all signatures. When the manual review result is inconsistent with the model judgment, the review result will be updated with a weight of one percent to update the reference distribution.

[0054] The remaining modules and parameters are the same as in Example 1.

[0055] Example 3 The differences between this embodiment and Embodiment 1 are as follows.

[0056] First, the sensor configuration was adjusted. The pen pressure sensor range was adjusted to 0 grams to 800 grams with an accuracy of ±3 grams. The accelerometer sampling frequency was adjusted to 250 Hz. The grip force sensor adopted a capacitive sensor with a sampling frequency of 200 Hz and a quantization accuracy of 14 bits.

[0057] Second, different implementations of the zero-knowledge proof protocol are selected. This embodiment still uses the Bulletproofs protocol, but parameter optimization makes the proof generation time no more than 1.2 seconds and the proof size no more than 0.8KB.

[0058] Third, the key rotation cycle has been adjusted to 18 months.

[0059] The remaining modules and parameters are the same as in Example 1.

[0060] Technical effect verification The above embodiments were tested and found to have a 0.00% success rate for signature forgery, a cumulative rejection rate of less than 3.1% for legitimate signatures within 6 months, a biological entropy Shannon entropy value of not less than 0.99, a health warning sensitivity of 92% to 96%, a specificity of 90% to 93%, a 100% success rate for switching downgrade modes, and a 100% success rate for key rotation. The test results fully demonstrate the advancements of the technical solution of the present invention in terms of security, usability, and functionality.

[0061] The accompanying drawings of the embodiments disclosed in this invention only involve structures related to the embodiments disclosed in this invention. Other structures can refer to general designs. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0062] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A medical blockchain-based evidence storage and AI-powered dynamic handwriting analysis system based on the national cryptographic standard SM2 signature. Includes, characterized in that: includes: The biological entropy generation module is configured to collect dynamic sequences of handwriting during the writing process, and the dynamic sequences of handwriting include at least time-series data of pen pressure, acceleration, and grip force. Feature vectors are extracted from the handwriting dynamic sequence, and the feature vectors are converted into 256-bit uniformly distributed biological entropy Ebio using the SHA-3 chaotic expansion function; the biological entropy Ebio and its hash commitment H(Ebio) are output. The SM2 signature random number mixing module is configured to obtain hardware true random number RTRNG; calculate the random number k of the SM2 signature algorithm = (RTRNG × OREbio) mod n, where n is the order of the SM2 elliptic curve. Use the random number k to generate an SM2 signature pair (r, s); Upload the signature pair (r,s) and the hash commitment H (Ebio) to the blockchain for evidence storage; The signature verification module is configured to obtain the signature pair (r, s) to be verified and the hash commitment H (Ebio) from the blockchain; Verify the mathematical correctness of the SM2 signature pair; At the same time, it verifies whether the hash commitment in this signature is consistent with the hash commitment recorded on the chain; A signature is considered valid only if both pass.

2. The medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature as described in claim 1, characterized in that: The biological entropy generation module also includes a degradation mode. When the data quality of the handwriting dynamic sequence is lower than a preset threshold, it automatically switches to a pure hardware true random number mode, that is, sets k=RTRNGmodn and records this degradation event on the blockchain.

3. The medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature as described in claim 1, characterized in that: The SM2 signature private key is managed using a (2,2) threshold secret sharing scheme. The first fragment of the private key is stored on the doctor's mobile terminal, and the second fragment is stored in the hardware security module. Generating a complete signature requires the participation of both shards; if either shard is missing, a valid signature cannot be generated.

4. The medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature as described in claim 1, characterized in that: It also includes a handwriting drift detection and adaptive threshold module, configured to construct a reference distribution using historical handwriting features within a 90-day rolling time window, including a mean vector μref and a covariance matrix Σref; Calculate the Mahalanobis distance Ddrift between the current signature handwriting feature vector xt and the reference distribution; dynamically adjust the judgment threshold of the signature verification module according to the value of the Mahalanobis distance Ddrift: when Ddrift≤1.5, use the standard threshold of 0.85; when 1.5<Ddrift≤2.5, record the drift log and keep the standard threshold unchanged; when 2.5<Ddrift≤3.5, adjust the signature verification threshold to 0.75 and trigger manual sampling review; when Ddrift>3.5, adjust the signature verification threshold to 0.70 and generate a doctor's health warning.

5. The medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature as described in claim 4, characterized in that: The method for generating the doctor's health warning is that the system uses a zero-knowledge scope proof protocol to generate a proof that Ddrift > 3.5 and the probability of neurodegenerative disease risk of handwriting feature mapping is greater than 0.

6. The zero-knowledge proof is stored on the blockchain, and the doctor can use their private key to decrypt and view detailed warning information.

6. The medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature as described in claim 4, characterized in that: The sampling rate for manual review is 1% to 5% of all signatures, and the reviewers make the judgment by comparing the pressure distribution map and the trajectory heat map. When the manual verification result is inconsistent with the model judgment, the verification result is updated with a weight of 1% to update the reference distribution.

7. The medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature as described in claim 1, characterized in that: The SHA-3 chaotic expansion function uses the SHAKE-256 algorithm. The input is a 128-bit original feature quantization value, and the output is a 256-bit uniformly distributed biological entropy Ebio. The Shannon entropy value of the output entropy is not less than 0.

99.

8. The medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature as described in claim 1, characterized in that: The SM2 signing private key adopts a periodic rotation mechanism, automatically generating a new SM2 key pair every two years. The old private key signs the new public key and is recorded on the chain as proof of key rotation. The old and new public keys are stored in the blockchain at the same time. New transactions are signed using the new private key, while historical transactions are still verified using the old public key.

9. The medical blockchain evidence storage and AI handwriting dynamic analysis system based on the national cryptographic SM2 signature as described in claim 1, characterized in that: The grip force data is collected by a piezoresistive or capacitive grip force sensor integrated into the smart pen, with a sampling frequency of not less than 100Hz and a quantization accuracy of not less than 12 bits. When the variance of grip force at 10 consecutive sampling points is lower than a preset threshold, the grip force signal is determined to be invalid, triggering the degradation mode described in claim 2.

Images