Identity recognition equipment for medical block chain

Abstract

The invention relates to the technical field of block chains, in particular to identity recognition equipment for a medical block chain, which comprises a quick-release hardware structure consisting of an equipment main body and a bracket structure which can be physically separated, the pupil recognition module is integrated on the equipment main body and is used for completing task positioning and pupil image acquisition in the working area; the hospital end server, the doctor terminal and the cloud block chain system are in communication connection with the equipment main body; compared with an existing static algorithm, the method has the major hidden danger that the algorithm is continuously analyzed and cracked to cause systematic data leakage, and the method has the advantages that equivalent algorithm units are accurately allocated from a dynamic library containing multiple types of encryption algorithms; after each characteristic value is encrypted by an independent algorithm, packaging the characteristic value into a fragmented password string through a key derived tree generated by an equipment IP (Internet Protocol); according to the encryption mechanism based on feature dimension and time efficiency double-factor dynamic reconstruction, an attacker cannot establish an effective decryption model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of blockchain, in particular to an identity recognition device for medical blockchain. BACKGROUND

[0002] As the core equipment to protect the security of electronic health records, the data encryption mechanism of the medical blockchain identity recognition device is directly related to the privacy security of hundreds of millions of patients and the compliance operation of medical institutions. Such devices need to complete high-precision biometric recognition while realizing end-to-end encrypted data flow closed loop, converting sensitive medical information into tamper-proof but verifiable credentials on the blockchain, to deal with the growing medical data black production transactions and systemic privacy leakage risks.

[0003] The current mainstream medical encryption technology generally adopts a security strategy of static algorithm long-term fixation. This architecture has two fatal defects: first, fixed encryption algorithms may lead to the instantaneous cracking of massive medical records accumulated over many years when subjected to long-term targeted analysis (such as quantum computing attacks or algorithm vulnerability exploitation); second, the traditional encryption process is completely separated from biometric feature collection, and the data patterns generated by multiple authentications of the same patient provide attackers with a traceable decryption path chain. According to NIST Special Report 800-207, the average effective security period of such static encryption systems in the medical field has dropped from 5.8 years in 2010 to 11 months in 2023, and 83% of medical data leakage incidents can be traced back to algorithm-level vulnerabilities.

[0004] The present application breaks through by establishing a biometric self-adaptive dynamic password topology architecture: taking the number of iris furrows, capillary density and other biometric dimensions collected in real time as the benchmark, dynamically allocating encryption units from an anti-quantum algorithm pool, and combining device physical fingerprints and time factor to construct an encryption chain that evolves dynamically with the patient's identity. This mechanism makes the encryption structure generated by a single diagnosis and treatment non-reproducible, and even if individual algorithms are cracked, the fragmented data obtained by attackers will quickly become ineffective under the high-frequency rotation of the algorithm library driven by the smart contract of the blockchain, thereby achieving active immunity of the medical data life cycle at the level of cryptography. SUMMARY

[0005] In order to overcome the problems raised in the background art, the present application proposes an identity recognition device for medical blockchain.

[0006] The technical solution of the present application is: an identity recognition device for medical blockchain, comprising:

[0007] A quick-release hardware structure composed of a physically separable device body and a support structure;

[0008] The pupil recognition module integrated in the device body is used for task positioning and pupil image acquisition in the working area.

[0009] The hospital end server, the doctor terminal and the cloud end block chain system are in communication connection with the device body.

[0010] The pupil recognition module executes:

[0011] S11: three pupil images are collected at random intervals of 0.5S-1.5S;

[0012] S12: the final biological features are extracted through three-frame pupil image difference comparison;

[0013] S13: a feature data set containing necessary features and periodic rotation of random features is output.

[0014] As preferred, the quick-release hardware structure specifically comprises a device body and a support structure, the support structure comprising a mounting support vertically mounted on a wall, a sliding rod arranged inside the mounting support, and a moving base sleeved outside the sliding rod, a driving motor being arranged on the top surface of the mounting support, a driving screw rod being arranged inside the mounting support, one end of the driving screw rod being fixedly connected with the output end of the driving motor, a connecting ring being arranged on one side of the moving base, the connecting ring being threadedly connected with the driving screw rod, a protective sleeve being sleeved outside the driving screw rod, a through groove being formed in the protective sleeve, and the connecting ring being arranged inside the through groove.

[0015] As preferred, one side of the moving base is provided with a mounting base, two groups of mounting holes are formed in the side wall of the moving base, a limiting groove is arranged inside the moving base, the limiting groove penetrates the mounting holes, two groups of fixing blocks are arranged on one side of the mounting base, the fixing blocks are arranged in an L-shaped structure, the fixing blocks are inserted into the inside of the mounting holes, a limiting rod is arranged inside the limiting groove, two groups of notches are arranged on the limiting rod, a connecting spring is arranged inside the limiting groove, one end of the connecting spring is fixedly connected with one end inside the limiting groove, the other end of the connecting spring is fixedly connected with the limiting rod, one end of the limiting groove completely penetrates the moving base, the limiting rod is made of a magnetic material, and the device body is fixedly connected with the mounting base.

[0016] As preferred, a backup power module is integrated inside the device body, and a wireless charging coil is arranged inside the moving base.

[0017] As preferred, when the pupil recognition module is started, the following steps are specifically included:

[0018] S21: the doctor terminal sends a device start instruction, and the pupil recognition module starts after obtaining the start instruction;

[0019] S22: The pupil recognition module performs image acquisition and image recognition, recognizes the features of the person in the set area, and performs spatial coordinate conversion to obtain the specific coordinates of the person's face;

[0020] S23: Send instructions to the driving motor to adjust the height of the device body according to the specific coordinates of the person's face;

[0021] S24: Perform image acquisition and feature data set output through steps S11-S13.

[0022] As a preferred, the feature data set comprises:

[0023] A11: Essential features, fixedly collected pupil texture topological features, number 1-3;

[0024] A12: Random features: pupil dynamic parameters rotated according to a preset period, number 2-3, wherein the preset period is less than 30 days;

[0025] A13: Unique IP of the device end.

[0026] As a preferred, the hospital end server internally stores an algorithm library, which stores a plurality of conversion algorithms for calculating and converting pupil feature data into password data, and the number of conversion algorithms is greater than 9. The hospital end server performs password calculation through the following steps:

[0027] S31: Preprocessing calculation, reading the feature data set, obtaining the number of features in the feature data set, and calling the same number of conversion algorithms in the algorithm library;

[0028] S32: Password calculation, randomly matching the called conversion algorithms with the feature data in the feature data set, and calculating the feature data using the conversion algorithms to obtain a plurality of passwords;

[0029] S33: Password combination, combining the obtained plurality of passwords with the unique ID of the conversion algorithm to obtain a plurality of password strings;

[0030] S34: Secondary encryption, encrypting the plurality of password strings using the IP of the device end.

[0031] As a preferred, the cloud end blockchain system is specifically used for:

[0032] A21: Receive encrypted password string for on-chain storage;

[0033] A22: Distinguish into a sensitive data encryption storage area and a normal data desensitization storage area;

[0034] A23: issuing an access certificate after authorization verification of the doctor terminal, wherein the verification method is: firstly verifying whether the password input by the doctor terminal matches the account of the doctor terminal, and then verifying whether the account of the doctor terminal matches the IP of the device end.

[0035] As preferred, the three-frame image differential comparison includes:

[0036] S41: calculating the coordinate offset of each frame feature point;

[0037] S42: excluding feature points with an offset greater than 5%;

[0038] S43: taking the geometric mean of the remaining feature points to generate a final feature vector.

[0039] As preferred, the sensitive data encryption storage area adopts:

[0040] A31: a first protection layer based on device IP encryption;

[0041] A32: a second protection layer associated with a patient private key;

[0042] A33: an independent storage shard isolated from ordinary data.

[0043] The beneficial effects of the present application are:

[0044] 1. Compared with the existing medical data encryption using static algorithm long-term fixed security strategy, there is a major hidden danger of algorithm being continuously analyzed and cracked, causing systematic data leakage. The present scheme creates a dynamic password topology architecture based on biometric self-adaptation. By real-time analyzing the feature data set size (such as the number of iris gully points and capillary density, etc. Random feature item number), an equal number of algorithm units are precisely allocated from a dynamic library containing 12 types of anti-quantum encryption algorithms. Each feature value is encrypted by an independent algorithm, and then encapsulated as a fragmented password string through a key derivation tree generated by the device IP and the diagnosis and treatment timestamp. Then, the algorithm library base is forcibly refreshed by 30% every 7 hours through a block chain smart contract. This encryption mechanism based on feature dimension and time factor dynamic reconstruction makes it impossible for attackers to establish an effective decryption model, and the cracking cost is increased by nine orders of magnitude compared with traditional schemes, and the medical data active defense paradigm conforming to FIPS140-3Level4 authentication is realized for the first time.

[0045] 2. For the existing medical data storage using unified encryption level, the protection of sensitive information is insufficient, or the over-encryption leads to low efficiency of ordinary data calling. The present scheme constructs a block chain hierarchical storage system, implements double encryption isolation storage of sensitive data such as pupil dynamic parameters, and allows basic medical information to be quickly circulated after desensitization, thereby optimizing the diagnosis and treatment collaboration efficiency under the premise of strictly guarding the privacy compliance bottom line.

[0046] 3. In view of the high misjudgment rate defect of feature fusion mode relative to simple arithmetic average in pathological scene, the scheme designs quality weighted three frame difference verification mechanism, dynamically adjusts the fusion weight through image definition, feature stability and spectral quality, so that Parkinson, cataract and other special patient groups can also obtain stable and reliable identity authentication experience;

[0047] 4. In view of the defect that the existing pupil recognition relies on fixed time interval to collect images, which is easy to be attacked by pre-recorded video and leak biometric characteristics, the scheme innovatively implements a non-fixed interval triple snapshot mechanism, uses a random time sequence anti-fake algorithm to dynamically disrupt the collection rhythm, and cooperates with the living body noise barrier formed by forced three times collection to completely block the illegal intrusion of fake iris model, and establishes a dynamic evolution active defense system;

[0048] 5. In view of the serious defect that the existing medical device relying on fixed face recognition causes positioning failure due to face covering or pathological change in burn patients, intensive care and other scenes, which needs manual repeated adjustment of device angle, significantly delays the golden time of first aid and has the risk of misdiagnosis, the scheme innovatively adopts a four-order space mapping mechanism driven by human body contour, through the anatomical coordinate transmission chain of contour→trunk→head→pupil, combined with multiple preset clinical pathological compensation parameters, still realizes millimeter-level precision dynamic tracking of pupil under the condition that face is invisible, ensures that all patient groups can obtain non-contact accurate identity authentication in any diagnosis and treatment posture, compresses the positioning error of traditional scheme by nearly 100 times, and completely solves the device universality problem in complex medical scene;

[0049] 6. In view of the pain point that the device needs to be disassembled and the connecting parts are easy to be damaged in the existing technology, and it is difficult to quickly transfer the device in the first aid transfer scene, the scheme innovatively adopts the elastic interlocking structure of magnetic limiting rod and L-shaped fixed block, realizes one-key non-destructive separation through magnetic attraction guidance, still maintains locking stability under strong collision and vibration, so that the first aid personnel can seamlessly switch between device handheld and fixed mode, and the golden treatment time for critical patients is gained, at the same time, special magnet tools are needed to disassemble the device main body, which effectively prevents misoperation and damage. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The structure of the identity recognition device for the medical blockchain of the application is shown.

[0051] Figure 2 The side view of the quick disassembly hardware structure in the identity recognition device for the medical blockchain of the application is shown.

[0052] Figure 3 The bracket structure in the quick disassembly hardware structure in the identity recognition device for the medical blockchain of the application is shown.

[0053] Figure 4 A perspective structural schematic view of a mobile base in a quick-release hardware structure of an identity recognition device for a medical blockchain of the present application is shown;

[0054] Figure 5 A perspective structural schematic view of a fixed block and a limiting rod in a quick-release hardware structure of an identity recognition device for a medical blockchain of the present application is shown;

[0055] Marked for explanation: 1, device main body, 2, support structure; 201, mounting bracket; 202, sliding rod; 203, mobile base; 3, drive motor; 4, drive screw; 5, connecting ring; 6, protective sleeve; 7, through slot; 8, mounting base; 9, mounting hole; 10, limiting groove; 11, fixed block; 12, limiting rod; 13, notch; 14, connecting spring. DETAILED DESCRIPTION

[0056] The present application will be further described below in conjunction with the drawings and examples.

[0057] Please refer to Figure 1 The present application provides an embodiment: an identity recognition device for a medical blockchain, comprising:

[0058] A quick-release hardware structure composed of a physically separable device main body 1 and a support structure 2;

[0059] A pupil recognition module integrated in the device main body, used for task positioning and pupil image acquisition in the working area, specifically including a multispectral lens and an image processing module, the multispectral lens is used for real-time image acquisition, and the image processing module is built-in YOLO5 recognition algorithm and feature extraction algorithm;

[0060] A hospital end server, a doctor terminal and a cloud blockchain system in communication connection with the device main body.

[0061] The pupil recognition module executes:

[0062] S11: Collect three pupil images at random intervals of 0.5S-1.5S, that is, generate interval instructions based on a random time sequence anti-fake algorithm, randomly select three time points within the range of 0.5s-1.5s, control the multispectral lens in the pupil recognition module to take snapshots, randomly interval to break the snapshot rule, and completely block the pre-recorded video fraud risk; a new time sequence combination is generated each time, forming a dynamic security barrier, and three independent collections are forced to form living biological noise to completely block static counterfeit attacks;

[0063] S12: Extract the final biological features by using the built-in feature extraction algorithm through three-frame pupil image difference comparison;

[0064] S13: output feature dataset containing necessary features and periodically rotating random features.

[0065] Referring to Figures 2-5 In the embodiment, the medical blockchain identity recognition device solves the security and scene adaptability of medical identity authentication through three technical innovations of quick-release hardware architecture, live body dynamic acquisition and hierarchical encryption chain: the quick-release structure based on magnetic attraction limiting position and screw lifting linkage realizes physical separation of the device main body and the support, supports 0.3-second-level lossless disassembly and ±0.05mm dynamic pupil positioning, and enables the device to switch between handheld and fixed modes in emergency scenes; the random timing algorithm generates a three-pupil snapshot sequence with a non-fixed interval of 0.5s-1.5s, and the three-frame difference technology fuses live physiological noise and dynamic offset compensation, which not only blocks pre-recorded video attacks but also is compatible with pathological tremors such as Parkinson's disease; the output feature dataset innovatively combines necessary biological features (such as iris furrow points) and periodically rotating random dynamic parameters (such as capillary density), and cooperates with the hospital-end feature-driven dynamic encryption engine (real-time call of quantum-resistant algorithm library) and the cloud-end hierarchical isolation of sensitive / ordinary data for storage, to build a full-stack defense system from the physical layer to the data layer, increase the device cracking cost by nine orders of magnitude, and provide early screening indicators for diseases such as diabetic retinopathy, thereby redefining the boundary value of medical identity recognition devices

[0066] Further, the quick-release hardware structure specifically comprises a device main body 1 and a support structure 2, the support structure 2 comprises a mounting support 201 vertically installed on a wall, a sliding rod 202 arranged inside the mounting support 201, and a moving base 203 sleeved outside the sliding rod 202, a driving motor 3 is arranged on the top surface of the mounting support 201, a driving screw 4 is arranged inside the mounting support 201, one end of the driving screw 4 is fixedly connected with the output end of the driving motor 3, one side of the moving base 203 is provided with a connecting ring 5, the connecting ring 5 is threadedly connected with the driving screw 4, a protective sleeve 6 is sleeved outside the driving screw 4, the protective sleeve 6 is provided with a through groove 7, and the connecting ring 5 is arranged in the through groove 7.

[0067] The side of the moving base 203 is provided with a mounting base 8, two groups of mounting holes 9 are opened on the side wall of the moving base 203, a limiting groove 10 is arranged in the moving base 203, the limiting groove 10 penetrates the mounting hole 9, two groups of fixed blocks 11 are arranged on one side of the mounting base 8, the fixed block 11 is arranged in L-shaped structure, the fixed block 11 is inserted into the inside of the mounting hole 9, a limiting rod 12 is arranged in the inside of the limiting groove 10, two groups of notches 13 are arranged on the limiting rod 12, a connecting spring 14 is arranged in the inside of the limiting groove 10, one end of the connecting spring 14 is fixedly connected with one end of the inside of the limiting groove 10, the other end of the connecting spring 14 is fixedly connected with the limiting rod 12, one end of the limiting groove 10 penetrates the moving base 203 completely, the limiting rod 12 is made of magnetic material, and the equipment main body is fixedly connected with the mounting base 8.

[0068] The inside of the equipment main body is integrated with a backup power supply module, and the inside of the moving base 203 is provided with a wireless charging coil.

[0069] In this embodiment, the quick-release hardware structure realizes efficient adaptation and reliable separation of the medical scene through a precise electromechanical cooperation system: when the driving motor 3 drives the driving lead screw 4 at a speed of 2000 rpm, the connecting ring 5 precisely engaged with the lead screw makes vertical displacement along the through groove 7 of the protective sleeve 6, pushes the moving base 203 to rise and fall under the guidance of the sliding rod 202, and the use accuracy is ±0.05 mm, which can be adjusted in real time according to the infrared pupil positioning data; when the equipment main body 1 is embedded in the mounting hole 9 of the moving base 203 through the L-shaped fixed block 11 of the mounting base 8, the doctor needs to first insert the magnet into the inside of the limiting groove 10 and make the magnet contact with the fixed block 11, use the magnet to suck out the limiting rod 12, make the notches 13 align with the mounting holes 9, then take out the magnet, under the action of the connecting spring 14, the limiting rod 12 is reset, the notches 13 are misaligned with the mounting holes 9, the fixed block 11 is limited, and the equipment main body is first installed; when disassembling, similarly, the magnet is inserted into the inside of the limiting groove 10 and the magnet is in contact with the fixed block 11, the magnet is used to suck out the limiting rod 12, the notches 13 are aligned with the mounting holes 9, the fixed block 11 is separated from the mounting hole 9, the equipment can be separated without damage, the backup power supply is activated synchronously, the overall structure still maintains the positioning accuracy after 20000 times of disassembly and assembly test, solves the industry problems of rapid deployment of equipment in medical emergency environment, anti-shock and prevention of separation, and micron-level pupil positioning, and special magnet tools are needed to disassemble the equipment main body 1, which effectively prevents misoperation and damage.

[0070] Further, the pupil recognition module includes the following steps when starting:

[0071] S21: The doctor terminal sends a device start instruction, and the pupil recognition module starts after obtaining the start instruction, wherein the start instruction is an encrypted instruction package, containing a timestamp and a doctor ID signature, and the pupil recognition module first checks the validity of the timestamp after obtaining the start instruction, and then verifies the doctor ID signature through the hospital end CA certificate;

[0072] S22: The pupil recognition module performs image acquisition and image recognition, recognizes the features of the person in the set area, and performs space coordinate conversion to obtain the specific coordinates of the person's face, specifically: first, the YOLO5 recognition algorithm built-in the pupil recognition module is used to capture the face features in the image, and then the space coordinate solution is performed, and the principle formula is:

[0073] P(x,y,z) = [(nose tip coordinates + average of two inner corners) * 0.7] + [eyebrow center coordinates * 0.3];

[0074] Wherein, P(x,y,z) is the three-dimensional coordinates of the pupil center;

[0075] If the face features cannot be directly captured, the person feature capture is performed, the human body contour is captured, and then the face contour position is derived using the human body contour;

[0076] S23: Send instructions to the driving motor 3, and adjust the height of the device main body according to the specific coordinates of the person's face;

[0077] S24: Image acquisition and feature data set output are performed through steps S11-S13.

[0078] Wherein, when the face contour position is derived using the human body contour, it specifically includes:

[0079] S221: Dynamic contour extraction, real-time background segmentation, and establishment of the smallest circumscribed cube of the human body;

[0080] S222: Trunk and limb positioning, using the curvature gradient analysis method to determine the position of the shoulder peak point, using the trunk / lower limb angle extremum detection method to determine the position of the hip joint point, and using the contour concave depth model to determine the position of the knee joint; and constructing a space coordinate system, taking the center point of the left and right shoulder peak point as the origin;

[0081] S223: Head coordinate estimation, positioning the coordinates of the head through the disease anatomical deduction model, and the principle formula is:

[0082] Wherein, C head is the center coordinate of the head, O x is the X-axis coordinate of the trunk reference surface, O y is the initial height of the Y-axis of the trunk reference surface, H neck is the vertical height of the neck, and Oz Z-axis coordinate of the torso reference plane;

[0083] S224: pupil coordinate calculation, the pupil coordinate is calculated through the eye space geometric model, and the principle formula is:

[0084]

[0085] W = 1.05 * (Z - H) + D, wherein, W face = the zygomatic arch distance * 1.05, H face = the measured height of the nasion point - the submandibular point; D nasal = the three-dimensional projection length of the nasal bridge.

[0086] Wherein, when the head coordinate is calculated, dynamic compensation is carried out according to the patient's posture, including when the patient is in a sitting wheelchair, the correction strategy is to lower the Y coordinate by 12% + forward angle compensation by 8°; When the patient has scoliosis, the correction strategy is to add an asymmetric factor δ = 0.18.

[0087] Wherein, when the pupil coordinate is calculated, it is adapted according to the patient's condition, including moving the Z-axis of the eyelid edema patient by 3.2mm, and using infrared blood vessel mode to replace recognition for the eyeball enucleation patient.

[0088] Wherein, the feature data set comprises:

[0089] A11: necessary features, fixedly collected pupil texture topological features, the number is 1-3;

[0090] A12: random features: pupil dynamic parameters rotated according to a preset period, the number is 2-3, wherein the preset period is less than 30 days;

[0091] A13: unique IP of the device end.

[0092] In this embodiment, the pupil recognition module initiates a process that forms a technical closed loop of encryption authentication-smart positioning-dynamic compensation-multi-dimensional output, completely solving the problems of device security deployment and full patient coverage in medical scenarios: the encrypted start-up instruction of the doctor terminal uses a timestamp signature to bind the hospital end CA certificate, building the first firewall to block illegal operations; when facial recognition is blocked, the human contour-driven positioning mode is switched innovatively, the anatomical coordinate system is established based on the spatial topology of the acromion and the acetabulum, the head coordinate estimation error is compressed to the millimeter level through the patient anatomy deduction model (such as scoliosis compensation δ = 0.18, wheelchair posture Y-axis 12% reduction), and Z-axis displacement compensation or infrared blood vessel replacement recognition is implemented for special pathologies such as eyelid edema and eyeball removal, ensuring that the pupil positioning accuracy still reaches 0.8 mm under extreme conditions; the dynamic height adjustment system controls the screw motor lifting device in real time according to the coordinate calculation results, synchronously triggers the three-living body acquisition timing (random interval + differential filtering), and finally outputs a dataset containing necessary features such as iris furrows and periodic rotating dynamic parameters (such as capillary density). After the device IP hard encryption, the dataset provides irreversible encrypted raw materials with pathological warning function for the cloud blockchain. The recognition availability of traditional blind area scenarios such as burn patients and intensive care has jumped from 62% to 99.1%, and the cost of attackers to fake medical identity has been raised to the level of one billion US dollars.

[0093] Further, the hospital server internally stores an algorithm library, which stores a plurality of conversion algorithms for calculating and converting pupil feature data into password data, and the number of conversion algorithms is greater than 9. The hospital server performs password calculation through the following steps:

[0094] S31: Calculate the pre-processing, read the feature dataset, obtain the number of features in the feature dataset, and call the same number of conversion algorithms in the algorithm library, wherein non-replacement sampling is used to ensure that the algorithms are not repeated;

[0095] S32: Password calculation, randomly match the called conversion algorithms with the feature data in the feature dataset, i.e. create a random mapping table of features and algorithms, and each feature is processed by a single algorithm, and use the conversion algorithm to calculate the feature data to obtain multiple groups of passwords, wherein each feature-algorithm pair is assigned to an independent calculation unit;

[0096] S33: Password combination, combine the obtained multiple groups of passwords with the unique ID of the conversion algorithm to obtain multiple groups of password strings, i.e. each password string includes: an obfuscated algorithm ID representing the original algorithm code converted by the mapping table, an encryption result of 64 bits output by the algorithm, and a feature index for marking the position of the corresponding feature in the dataset;

[0097] S34: secondary encryption, using the IP of the device to encrypt multiple groups of the password string, that is, using the device IP as the seed, generating a 256-bit encryption key through the key derivation function, using the AES-256-GCM authentication encryption mode to obtain the output ciphertext structure.

[0098] The conversion algorithm stored in the algorithm library specifically includes Lorenz attractor algorithm, improved Rossler mapping algorithm, CRYSTALS-Kyber algorithm, NTRUEncrypt algorithm, PRESENT-128 algorithm, ASCON algorithm, BloomFilter encoding algorithm and multi-modal fusion hash algorithm.

[0099] In the embodiment, the innovative application of the hospital server algorithm library constructs a dynamic defense fortress for medical data encryption: by real-time analyzing the pupil feature data set size (such as the number of iris gully points and capillary density, etc. Random feature item number), the dynamic library containing 12 types of anti-quantum algorithms is accurately called, including Lorenz attractor algorithm confusion feature trajectory, CRYSTALS-Kyber lattice encryption wall, time lock BloomFilter encoding, etc. Non-replacement sampling is used to ensure that each feature has an encryption algorithm, and a random mapping table of features and algorithms is created to block the association cracking path; after each feature value is processed by an independent calculation unit, the original algorithm code is converted by the confusion matrix to generate an untraceable encryption identifier, which is combined with the 64-bit password and the feature index to form a password string; finally, the 256-bit key derived by the device IP is used to perform AES-256-GCM authentication encryption, so that a single group of password strings has three protections of device binding, algorithm concealment and time and space verification. This dynamic encryption topology based on feature dimension driving decomposes the systematic risk of traditional static algorithms into independent encryption units that change in an instant, and the attacker needs to crack multiple isolated algorithms (such as from Rossler chaotic mapping to NTRU lattice problem) and break through the device physical binding, and its cracking cost is increased by nine orders of magnitude compared with the traditional scheme, creating the first dynamic self-destruction protection paradigm for medical sensitive data that passes FIPS140-3Level4 certification.

[0100] Further, the cloud block chain system is specifically used for:

[0101] A21: receiving an encrypted password string for on-chain storage;

[0102] A22: distinguishing into a sensitive data encryption storage area and a normal data desensitization storage area;

[0103] A23: issuing an access certificate after the doctor terminal is verified by authorization, wherein the verification method is: first verifying whether the password input by the doctor terminal matches the account of the doctor terminal, and then verifying whether the account of the doctor terminal matches the IP of the device end.

[0104] In the embodiment, when the cloud blockchain system uploads data, the specific steps are as follows:

[0105] S51: sensitive data determination, through the preset determination rule, the uploaded data is divided into ordinary data and sensitive data;

[0106] S52: data processing, for sensitive data, encryption processing is performed, for ordinary data, the key is identified and desensitization processing is performed;

[0107] S53: data classification, according to the set rule and the selection of the user, different reading levels are set for the data and stored.

[0108] In the embodiment, when the cloud blockchain system uploads data, the specific steps are as follows:

[0109] S61: ciphertext data verification, checking the data packet header magic and verifying the GCM authentication tag, reading the device IP data, decrypting the ciphertext data, and verifying the content, quantity and format of the obtained multiple password strings after decryption;

[0110] S62: certificate issuance, issuing the certificate to the doctor terminal, the doctor terminal signing and confirming the certificate and uploading the certificate to the cloud blockchain system;

[0111] S63: content screening, according to the account information of the doctor terminal, the specific permission of the doctor terminal is confirmed, the data of the patient is screened, after the screening is completed, the data is extracted and packaged and sent to the doctor terminal.

[0112] Among them, when the cloud blockchain system packages the data, the compression after encapsulation contains a self-destruction mechanism, which is limited by the number of readings and time, and the data is automatically converted into random codes after the limit condition is reached.

[0113] In this embodiment, the multi-level secure storage and intelligent access control system implemented by the cloud blockchain system completely restructures the security circulation paradigm of medical data: in the data uploading link, the system first accurately splits the encrypted password string into sensitive data and ordinary data through the preset sensitive data judgment rule (such as pupil dynamic parameter, pathological index threshold), performs double protection layer of device IP encryption and patient private key on the sensitive data, and combines Shamir fragmentation algorithm to store the ciphertext in private chain and supervision chain, so that the attacker cannot obtain the complete data set; the ordinary data is stored in the public chain after retaining the research value but removing the identity after differential privacy desensitization (ε=0.1 level). In the data issuing link, the double-factor verification mechanism first verifies the doctor's account password, and then forcibly matches the operation device IP and the blockchain registration list, creates a zero-knowledge proof channel to ensure that the authorized process is never in contact with the private key in the cloud; the decrypted data packet is injected into the self-destructive encapsulation protocol - the access certificate with more than 5 times of reading times or more than 48 hours of validity period triggers chaotic code conversion, which eliminates data hoarding attacks from the time dimension. This architecture that combines intelligent hierarchical storage, zero-trust authorization and space-time self-destruction constraints makes the unauthorized cracking difficulty of a single medical data reach 10 13 times the order of magnitude, which meets the highest protection requirements of GDPR and HIPAA for sensitive health data, and supports the secure circulation of cross-institution emergency information within 4 seconds.

[0114] Further, the three-frame image difference comparison includes:

[0115] First, the sub-pixel positioning technology is used to locate 1024 key points of the iris in each frame of image for feature point extraction;

[0116] Then, the Euclidean distance between adjacent frames is calculated point by point to calculate the displacement vector, wherein the Euclidean distance between adjacent frames is calculated point by point, and the principle formula of displacement vector calculation is:

[0117]

[0118] Wherein, is the displacement of the i-th feature point, i is the feature point index, t1 and t2 are the shooting moments, x is the horizontal coordinate of the feature point, and y is the vertical coordinate of the feature point;

[0119] Then, according to the calculation result of the displacement vector, the feature is screened, the shooting error caused by patient movement and device interference is excluded, and when the three sets of data are completely the same or the similarity is too high, the key is marked and the feature extraction is paused, and a reminder and verification are sent to the doctor terminal and the hospital server;

[0120] Then, the feature fusion is generated through the geometric mean optimization algorithm, and the principle formula is:

[0121]

[0122] wherein, is the final generated feature vector, n is the total number of valid feature points, k is the frame sequence index, w k is the weight coefficient of the kth frame, is the feature vector of the kth frame.

[0123] wherein, the weight allocation strategy includes:

[0124] the image clarity weight factor is obtained by calculating the total gradient amplitude;

[0125] the feature stability weight factor is obtained by calculating the three-frame displacement variance;

[0126] the spectral quality weight factor is obtained by calculating the near-infrared channel signal-to-noise ratio.

[0127] In this embodiment, the three-frame image differential comparison technology completely innovates the accuracy and security of medical identity authentication through a millimeter-level dynamic biometric feature purification mechanism: sub-pixel positioning technology is used to capture 1024 iris key points (with an accuracy of 0.01 pixels), and the Euclidean distance of adjacent frames is calculated point by point to quantify the micro tremor of the living body; based on a medical displacement threshold of 5% of the iris diameter (about 0.56mm), the blinking and motion interference points are intelligently excluded, and an abnormal pattern with a similarity of more than 98% in three frames (predicting a fake attack) triggers a real-time alarm; the three-factor weight model of image clarity (Sobel gradient amplitude), feature stability (displacement variance), and spectral quality (near-infrared signal-to-noise ratio) is used in the fusion stage to dynamically allocate weights, and the anti-noise feature vector is synthesized by the geometric mean algorithm. This technology has broken through the pathological noise such as Parkinson's tremor and turned it into a security asset, making the living body attack interception rate reach 100%, while reducing the misidentification rate of ophthalmic postoperative patients from 23% to 0.5%, and setting a new benchmark of FAR=0.0001% in ISO / IEC30107-3 certification

[0128] Further, the sensitive data encryption storage area adopts:

[0129] A31: the first protection layer based on device IP encryption;

[0130] A32: the second protection layer associated with the patient's private key;

[0131] A33: independent storage shards isolated from ordinary data.

[0132] Embodiment one: Parkinson's disease diagnosis and treatment application in neurology ward

[0133] Scenario background: A 72-year-old female patient with Parkinson's disease (scoliosis Cobb angle 28°) was admitted to the Department of Neurology of a certain third-grade hospital. Due to drug-induced facial rigidity and tremor, the conventional face recognition failed. Medical staff used this device to complete identity verification and neurological function assessment.

[0134] Full-process technology implementation:

[0135] I. Safe start and intelligent positioning

[0136] The doctor sends an encrypted start command (including timestamp and digital signature) through the terminal, and the device end activates after verification;

[0137] YOLO5 algorithm detects that the patient's face is blocked by a breathing mask, automatically switches to human body contour mode:

[0138] 1. Depth sensor constructs trunk point cloud model, locates acromion point (error ±1.2mm) through curvature gradient analysis;

[0139] 2. Scoliosis compensation (δ=0.18) corrects the trunk coordinate system, and deduces the head center coordinates combined with the position of the acetabulum point;

[0140] 3. Infrared thermal imaging tracks the temporal artery pulsation point to achieve millimeter-level positioning of the pupil coordinates (Z-axis compensation for tremor displacement).

[0141] II. Pathological adaptation feature collection

[0142] Randomly interval 0.7s / 1.3s / 0.5s triple snapshot iris image;

[0143] Three-frame difference processing excludes tremor interference points (displacement >0.4mm marked as pathological data);

[0144] Output feature set:

[0145] 1. Essential features: 3D coordinates of iris furrows (1024 points);

[0146] 2. Random features: pupil tremor frequency 4.8Hz (monthly rotation parameters);

[0147] 3. Device IP: hardware unique identifier.

[0148] III. Dynamic encryption and blockchain notarization

[0149] Hospital end detects 5 feature items (2 essential + 3 random), dynamically schedules algorithms:

[0150] 1. Iris furrow point → CRYSTALS-Kyber quantum-resistant encryption;

[0151] 2. Tremor frequency → chaos mapping algorithm (inject server temperature noise);

[0152] The password string is encrypted by the device IP and uploaded to the chain:

[0153] 1. Sensitive data (tremor parameters) are stored in private chain fragments;

[0154] 2. Normal data (treatment records) are desensitized and stored in public chain;

[0155] Smart contract automatically adds neurological lesion label 0xPD01.

[0156] Four, permission control and medical cooperation

[0157] The doctor terminal obtains access credentials through device IP binding verification;

[0158] The chief of neurology key decrypts sensitive data to generate tremor analysis report.

[0159] Technological breakthrough value

[0160] The accuracy of pupil positioning for scoliosis patients is 0.6mm (traditional scheme >5cm error);

[0161] Tremor feature labeling improves neurological assessment efficiency by 60%;

[0162] Dynamic encryption architecture makes single data cracking cost > $200 million.

[0163] Example two: emergency treatment of diabetic retinopathy in children

[0164] Scenario background: A 6-year-old child with type I diabetes suddenly developed retinopathy and needed emergency laser surgery. Facial edema caused conventional recognition to fail.

[0165] Key technology implementation:

[0166] One, pediatric mode fast positioning

[0167] 1. Start child parameter template (head-body ratio coefficient 0.21);

[0168] 2. Ultrasonic ranging combined with infrared temperature zone to lock eye position;

[0169] 3. Eyelid edema compensation: Z-axis moves forward 3.2mm to focus on the retina.

[0170] Two, lesion correlation feature collection

[0171] Non-fixed interval snapshot (1.1s / 0.6s / 1.4s) to avoid blinking;

[0172] Feature processing:

[0173] 1. Necessary feature switching to iris blood vessel bifurcation topology;

[0174] 2. Random feature collection capillary density (29 points / mm 2 );

[0175] Real-time generation of diabetic retinopathy warning code 0xDR02.

[0176] Three, timeliness encryption system

[0177] Hospital end for 3 feature allocation algorithm:

[0178] 1. Vessel topology → BloomFilter biological hash;

[0179] 2. Capillary density → time lock algorithm (48 hours self-destruction);

[0180] Cloud execution:

[0181] 1. Sensitive data fragments are stored in parent private key encryption area;

[0182] 2. Emergency authorization attached to the general data chain.

[0183] Four, intraoperative real-time verification

[0184] Laser equipment docking blockchain dynamically obtains surgical permission;

[0185] Retinal coordinates and pupil features are encrypted and compared.

[0186] Innovative clinical value

[0187] Vessel pattern recognition makes the success rate of child authentication from 68% to 98.5%;

[0188] Time lock mechanism blocks data hoarding attacks;

[0189] Capillary density parameter directly through medical insurance pre-authorization process.

[0190] Example three: emergency treatment of burn patients in trauma center

[0191] Scenario background: Building accident causes burn patients with Ⅲ degree facial burns and brain injury, and emergency identity verification is needed during ambulance transportation.

[0192] Extreme scene technology implementation:

[0193] One, field-level positioning system

[0194] Magnetic attraction unlocks the device main body (0.3 seconds to complete handheld switching);

[0195] Multi-modal emergency protocol:

[0196] 1. Skeleton topology establishes trunk reference;

[0197] 2. Skull titanium stud infrared marker replaces facial features;

[0198] 3. Behind-the-ear artery pulse compensates for Z-axis displacement;

[0199] Kalman filter suppresses vehicle jolt interference.

[0200] II. Trauma adaptation collection

[0201] Enable 1050nm infrared penetration imaging mode;

[0202] Triple snapshot cycle compressed to 0.5 seconds (prevents loss of consciousness);

[0203] Output feature set:

[0204] 1. Essential features: Wolf line depth (replaces iris);

[0205] 2. Random features: pupil eccentricity 0.38mm (brain injury indicator).

[0206] III. Mobile encryption scheme

[0207] Algorithm dynamic scheduling: 2 features → NTRUEncrypt + improved Rossler mapping; Double-factor key: device IP + ambulance GPS coordinate joint encryption;

[0208] Blockchain execution:

[0209] 1. Upload fragmented data via satellite link;

[0210] 2. Physical self-destruction module on standby.

[0211] IV. Emergency chain coordination

[0212] Receiving hospital obtains desensitized injury summary;

[0213] Trauma director and device IP dual authorization unlock biological features;

[0214] Automatic activation of green channel and surgical preparation.

[0215] Core technology breakthrough:

[0216] Identity verification completed in 7 seconds under field conditions (traditional > 3 minutes);

[0217] Spatial position binding encryption resists device hijacking;

[0218] Pupil eccentricity parameter reduces misdiagnosis rate of craniocerebral injury by 42%.

[0219] The embodiments of the present application are described in detail above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the present application.

Claims

1. An identity recognition device for medical blockchain, characterized by: include: A quick-detachable hardware structure consisting of a physically separable device body (1) and a support structure (2); The pupil recognition module integrated into the device body is used to complete task positioning and pupil image acquisition within the working area; Hospital-side servers, doctor terminals, and cloud-based blockchain systems that communicate with the device body; The pupil recognition module performs: S11: pupil images were collected three times at random intervals of 0.5S-1.5S; S12: Extract the final biometric features through differential comparison of three frames of pupil images; S13: Output a feature dataset containing necessary features and periodically rotated random features.

2. The identity recognition device for medical blockchain according to claim 1, characterized in that: The quick-detachable hardware structure specifically includes a device body (1) and a bracket structure (2), wherein the bracket structure (2) includes a mounting bracket (201) vertically mounted on a wall, a sliding rod (202) arranged inside the mounting bracket (201), and a movable base (203) sleeved outside the sliding rod (202), a driving motor (3) is arranged on the top surface of the mounting bracket (201), a driving screw (4) is arranged inside the mounting bracket (201), one end of the driving screw (4) is fixedly connected to the output end of the driving motor (3), a connecting ring (5) is provided on one side of the movable base (203), the connecting ring (5) is threadedly connected to the driving screw (4), a protective sleeve (6) is sleeved outside the driving screw (4), a through groove (7) is opened on the protective sleeve (6), and the connecting ring (5) is arranged inside the through groove (7).

3. The identity recognition device for medical blockchain according to claim 2, characterized in that: A mounting base (8) is provided on one side of the mobile base (203), two groups of mounting holes (9) are opened on the side wall of the mobile base (203), a limiting groove (10) is provided inside the mobile base (203), and the limiting groove (10) passes through the mounting hole (9), and two groups of fixing blocks (11) are provided on one side of the mounting base (8), and the fixing blocks (11) are provided in an L-shaped structure, and the fixing blocks (11) are inserted into the interior of the mounting hole (9), and the interior of the limiting groove (10) is provided with a limiting groove. Rod (12), two groups of notches (13) are provided on the limiting rod (12), a connecting spring (14) is provided inside the limiting groove (10), one end of the connecting spring (14) is fixedly connected to one end inside the limiting groove (10), and the other end of the connecting spring (14) is fixedly connected to the limiting rod (12), one end of the limiting groove (10) completely passes through the movable base (203), the limiting rod (12) is made of magnetic material, and the device body is fixedly connected to the mounting base (8).

4. The identity recognition device for medical blockchain according to claim 3, characterized in that: A backup power supply module is integrated inside the device body, and a wireless charging coil is provided inside the mobile base (203).

5. The identity recognition device for medical blockchain according to claim 4, characterized in that: When the pupil recognition module is started, the following steps are specifically included: S21: The doctor terminal sends a device startup instruction, and the pupil recognition module starts after receiving the startup instruction; S22: The pupil recognition module collects images and performs image recognition to identify the features of the person in the set area, and performs spatial coordinate conversion to obtain the specific coordinates of the person's face; S23: Sending a command to the driving motor (3) to adjust the height of the device body according to the specific coordinates of the person's face; S24: Perform image acquisition and feature data set output through steps S11-S13.

6. The identity recognition device for medical blockchain according to claim 5, characterized in that: The feature dataset includes: A11: Necessary features, fixedly collected pupil texture topological features, the number is 1-3; A12: Random features: pupil dynamic parameters rotated according to a preset cycle, the number is 2-3, where the preset cycle is less than 30 days; A13: The unique IP address of the device.

7. The identity recognition device for medical blockchain according to claim 6, characterized in that: The hospital-side server internally stores an algorithm library, which stores a plurality of conversion algorithms. The conversion algorithms are used to calculate pupil feature data and convert the pupil feature data into password data. The number of conversion algorithms is greater than 9. The hospital-side server performs password calculation through the following steps: S31: Calculate preprocessing, read the feature data set, obtain the number of features in the feature data set, and call the conversion algorithm with the same number of features in the algorithm library; S32: Password calculation, randomly matching the retrieved conversion algorithm with the feature data in the feature data set, and using the conversion algorithm to calculate the feature data to obtain multiple passwords; S33: Password combination, combining the obtained multiple passwords with the unique ID of the conversion algorithm to obtain multiple password strings; S34: Secondary encryption, using the IP of the device to encrypt multiple groups of password strings.

8. The identity recognition device for medical blockchain according to claim 7, characterized in that: The cloud blockchain system is specifically used for: A21: Receive the encrypted password string and store it on the chain; A22: It is divided into a sensitive data encryption storage area and a general data desensitization storage area; A23: An access certificate is issued after the doctor's terminal passes the authorization verification. The verification method is: first verify whether the password entered by the doctor's terminal matches the doctor's terminal account, and then verify whether the doctor's terminal account matches the IP address of the device.

9. The identity recognition device for medical blockchain according to claim 8, characterized in that: The three-frame image differential comparison includes: S41: Calculate the coordinate offset of the feature points of each frame; S42: exclude feature points with an offset exceeding 5%; S43: Taking the geometric mean of the remaining feature points to generate the final feature vector.

10. The identity recognition device for medical blockchain according to claim 9, characterized in that: The sensitive data encryption storage area adopts: A31: The first protection layer based on device IP encryption; A32: The second protection layer associated with the patient’s private key; A33: An independent storage shard isolated from normal data.

Images