Face data full life cycle management system and method

Abstract

The invention belongs to the technical field of data security and privacy protection, and particularly relates to a face data full-life-cycle management system which comprises a full-life-cycle management system. A data acquisition module, a data storage module, a data processing module, a data distribution module, a data destruction module and a supervision and analysis module are arranged in the full life cycle management system; the invention further particularly relates to a human face data full-life-cycle management method. The method comprises the following specific steps of personal authorization, human face collection and verification, safe storage and distribution, authorization access control and full-life-cycle supervision and analysis. According to the invention, through comprehensive management of personal authorization, face collection, identity and quality verification, state secret storage and distribution, revocation and destruction, traceability and other links of face data, full-life-cycle supervision of the face data of colleges and universities from generation to destruction is realized, and the safety of the face data and personal information is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of data security and privacy protection, and specifically relates to a facial data full lifecycle management system and method. Background Art

[0002] Facial recognition technology is currently widely used by colleges and universities in scenarios such as school gates, dormitories, canteens, unified identity authentication, and new student identity verification, and has been well received by teachers and students. However, network security, data security, and personal information security also need to be paid attention to.

[0003] Problems with existing technologies:

[0004] Difficulty in management: Unable to uniformly manage teacher and student facial recognition. Different brands, manufacturers, usage scenarios, and facial quality requirements lead to repeated collection of teacher and student facial information, increasing management difficulty, reducing user satisfaction, and increasing the risk of leakage.

[0005] Imperfect authorization mechanism: Individuals cannot authorize or withdraw facial recognition authorization, making closed-loop management impossible.

[0006] Insecure data storage and distribution mechanisms: The lack of secure and compliant data storage and distribution mechanisms leads to the disorderly use of facial data and personal sensitive information, posing a risk of leakage.

[0007] Passive supervision: It is impossible to manage the facial data of all teachers and students throughout their entire life cycle, and to provide unified management, supervision, early warning, and analysis, resulting in a relatively passive work process.

[0008] Compliance risk: If facial recognition is mandatory, complaints may occur, causing adverse effects. Summary of the Invention

[0009] The purpose of the present invention is to provide a facial data full life cycle management system and method, which can realize the full life cycle supervision of the facial data of college teachers and students from generation to destruction, and ensure the security of facial data and personal information.

[0010] The technical solutions adopted by the present invention are as follows:

[0011] A facial data full lifecycle management system, comprising a full lifecycle management system, wherein the full lifecycle management system is internally provided with a data acquisition module, a data storage module, a data processing module, a data distribution module, a data destruction module, and a supervision and analysis module;

[0012] The data storage module uses the national secret encryption algorithm SM to encrypt and store the data in the face database; during the data transmission process, the SSL protocol is used to encrypt the data;

[0013] The data distribution module uses the SSL / TLS protocol to encrypt the transmission channel; when establishing a connection, the system will conduct strict identity authentication, and only authorized business systems can obtain face copy data; at the same time, the integrity of the transmitted data is checked;

[0014] The data distribution module uses implicit watermark embedding and traceability to perform implicit watermark processing when sending it to the face business system, and provides real data traceability capabilities to trace the source of facial data in the event of security incidents such as data leakage;

[0015] The data distribution module adopts an authorized access control mechanism: a role-based access control mechanism is established to assign different access rights to different business systems, ensuring that facial data is only used within the authorized scope;

[0016] The data destruction module adopts cross-system data destruction: when an individual withdraws the authorization to use facial data, the system first destroys the user's facial data in this system; then, it sends a destruction instruction to all business systems that obtain the user's facial data through a predefined interface; after receiving the instruction, the business system immediately deletes the relevant data stored locally and feeds back the destruction result to this system.

[0017] The data acquisition module is used to enable the system to perform quality assessment on the collected images in real time during the collection process of the mobile phone camera.

[0018] The data processing module adopts a quality verification algorithm: the quality of the collected facial images is verified. The algorithm comprehensively considers the characteristic indicators of the image. Only when the various image quality indicators reach the preset threshold, the image quality is considered to be qualified.

[0019] The data processing module adopts a feature extraction algorithm: feature extraction is performed on a face image of qualified quality, and the extracted feature values ​​have high discrimination and stability, and are used to accurately represent the biological features of the face.

[0020] The supervision analysis module includes a supervision log recording unit. The system records the entire life cycle of facial data in detail, including data collection time, collection personnel, authorized business system, destruction time and other information.

[0021] The supervision and analysis module includes a real-time monitoring unit, through which supervisors can view the usage of facial data in real time through the system interface, and also has abnormal access detection.

[0022] The supervision analysis module includes a data analysis and report generation unit. The system regularly analyzes supervision logs and generates various statistical reports.

[0023] A method for managing the entire life cycle of facial data, including the following specific steps:

[0024] Step 1: Personal Authorization: Before facial data collection, the system displays an authorization agreement to the individual, clearly informing them of the purpose of collection, scope of use, storage period, and other information about their personal data. The individual then authorizes the operation through the system interface.

[0025] Step 2: Face collection and verification: Facial images are collected using mobile phone cameras according to unified standards and procedures. The collected facial images are then verified using a quality assessment algorithm to ensure they are clear, unobstructed, and free of blur. At the same time, the collected facial images are linked to personal identification information for public security facial identity verification.

[0026] Step 3: Secure storage and distribution: The verified facial data is encrypted using a national encryption algorithm and stored in the database. When other business systems need to use facial data, the system distributes the matching data to the relevant business systems via SSL secure encryption.

[0027] Step 4: Authorized access control: Establish a role-based access control mechanism to assign different access rights to different business systems, ensuring that facial data is only used within the authorized scope;

[0028] Step 5: Full life cycle supervision analysis: The system conducts full life cycle supervision analysis on the user's facial data from generation to destruction, records information such as data collection time, verification status, usage, destruction time, and generates a full supervision log; supervisors view the supervision log through the system interface and conduct real-time monitoring and analysis of the use of facial data.

[0029] In step 3, add the face tracing function. The specific steps are as follows:

[0030] Step 3.1: Watermark information generation and encryption: Construct the traceability mark and embed the watermark fragment;

[0031] Step 3.2: Data distribution and watermark association: Construct a watermark-distribution record mapping table and store it in encrypted form;

[0032] Step 3.3: When data leakage occurs, perform implicit watermark tracing;

[0033] Step 3.4: Watermark extraction and decoding: feature vector sampling and logo reconstruction;

[0034] Step 3.5: Provenance Verification: Record query and matching, and perform the query in the watermark-distribution record mapping table based on the reconstruction identifier.

[0035] In step 5, the facial data is revoked and destroyed. The specific steps are as follows:

[0036] Step 5.1: Trigger conditions and event modeling: User triggers the authorization withdrawal event and data status mark;

[0037] Step 5.2: Destroy the data in this system: perform logical deletion and physical removal;

[0038] Step 5.3: Cross-system command broadcast: pre-define the destruction command and use the broadcast mechanism to send the command to all related business systems;

[0039] Step 5.4: Business system response and feedback;

[0040] Step 5.5: State synchronization and auditing: The main system verifies the response signatures of all business systems, updates the global state, and writes to the audit log.

[0041] The technical effects achieved by the present invention are:

[0042] The present invention comprehensively manages the personal authorization, facial collection, identity and quality verification, national secret storage and distribution, revocation and destruction, and traceability of facial data, thereby achieving full life cycle supervision of the facial data of university teachers and students from generation to destruction, and ensuring the security of facial data and personal information.

[0043] This invention proposes for the first time the concept of full life cycle management of university facial data, covering all aspects from data collection to destruction, and realizing comprehensive supervision of facial data.

[0044] The present invention unifies multiple algorithm standards and can adapt to various brands of face recognition quality algorithms, such as Megvii, SenseTime, Baidu, etc., solving the problem of compatibility of quality standards between different algorithms.

[0045] The face tracing function of the present invention is to stamp the issued negative film with an implicit digital watermark. When a security incident such as data leakage occurs, the system can trace the source of the data through the digital watermark information and determine the responsible party.

[0046] The facial data revocation and destruction function of the present invention allows for the system to quickly and uniformly destroy relevant data from this system to all business systems that have obtained the user's facial data, using cross-system data destruction technology when an individual withdraws authorization to use facial data. This also ensures that the data cannot be recovered during the destruction process. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a system diagram of a management system provided by an embodiment of the present invention;

[0048] Figure 2 is a flow chart of a management method provided by an embodiment of the present invention;

[0049] Figure 3is a flow chart of the face tracing function provided by an embodiment of the present invention;

[0050] Figure 4 This is a flow chart of data revocation and destruction provided by an embodiment of the present invention.

[0051] In the accompanying drawings, the components represented by the reference numerals are as follows:

[0052] 1. Full life cycle management system; 101. Data acquisition module; 102. Data storage module; 103. Data processing module; 104. Data distribution module; 105. Data destruction module; 106. Supervision and analysis module. DETAILED DESCRIPTION

[0053] In order to make the purpose and advantages of the present invention more clearly understood, the present invention is described in detail below with reference to the following examples. It should be understood that the following text is only used to describe one or more specific embodiments of the present invention and does not strictly limit the scope of protection of the present invention.

[0054] like Figure 1 As shown, a facial data full life cycle management system includes a full life cycle management system 1, and the full life cycle management system 1 is internally provided with a data acquisition module 101, a data storage module 102, a data processing module 103, a data distribution module 104, a data destruction module 105 and a supervision and analysis module 106.

[0055] (1) data acquisition module 101;

[0056] During the mobile phone camera acquisition process, the system conducts real-time quality assessment of the acquired images, such as testing image brightness, contrast, clarity and other indicators. If the image quality does not meet the standards, the user is prompted to re-acquire the image.

[0057] The specific process is as follows:

[0058] Step 1: Brightness detection

[0059] Mathematical formula: Where I(x i ,y i ) represents the pixel (x i ,y i ) grayscale value, N is the total number of pixels, and the brightness threshold can be set as [L min , L max ], if it exceeds the range, it is judged that the brightness does not meet the standard;

[0060] Step 2: Contrast Assessment

[0061] Mathematical formula: The contrast is measured by calculating the standard deviation of the pixel value and the average brightness. If C RMS <C threshold , indicating insufficient contrast;

[0062] Step 3: Clarity detection

[0063] Tenengrad gradient method: Where Gx and Gy are the Sobel operator convolution results of the image in the horizontal and vertical directions respectively, and the clarity threshold S threshold Need to be adjusted dynamically according to equipment performance;

[0064] Step 4: Multiple indicators joint judgment

[0065] The system uses preset thresholds to perform a weighted comprehensive score on brightness, contrast, and clarity:

[0066] Among them, w1, w2, w3 are weight coefficients, the sum of which is 1. If Q < Q threshold , triggering a re-sampling prompt;

[0067] Step 5: Real-time feedback optimization

[0068] Using dynamic data management logic, when the system detects quality problems, it displays specific indicators (such as "insufficient brightness" or "blurred image") through the mobile phone interface and guides users to adjust the shooting angle or ambient light source.

[0069] According to the above content, the data acquisition module quantifies image quality indicators through mathematical modeling and combines dynamic feedback mechanism to achieve efficient acquisition, which meets the requirements of university management system for data accuracy and real-time performance.

[0070] (2) data storage module 102;

[0071] Encrypted storage: The national secret encryption algorithm SM4 is used to encrypt and store the data in the face database; during the data transmission process, SSL is used to encrypt the data to ensure the security of the data during transmission;

[0072] 2.1) SM4 encryption algorithm uses block cipher mode. The core steps can be divided into:

[0073] Step 1, key expansion: expand the initial 128-bit key into 32 round keys,

[0074] Where CKi is a fixed parameter and T′ is a nonlinear transformation function;

[0075] Step 2: Round function operation: each round performs a nonlinear transformation on a 128-bit data block.

[0076] The T function contains the S-box permutation (four 8-bit input nonlinear tables) and the linear transformation L, where L is:

[0077]

[0078] 2.2) Encrypted storage implementation process:

[0079] Step 1: Preprocessing of facial feature vectors: Split the 128-dimensional facial feature vector into blocks according to the SM4 group length (128 bits), and fill the missing parts to complete blocks;

[0080] Step 2, CBC mode encryption: Use the initialization vector (IV) to achieve inter-block correlation and prevent repeated plaintext from generating the same ciphertext: Where, C0=IV,P i is the i-th plaintext block, C i is the corresponding ciphertext block.

[0081] 2.3) Data transmission encryption is implemented using the SSL protocol:

[0082] Step 1, key exchange and authentication:

[0083] ECDH key agreement (based on elliptic curve): Both parties choose elliptic curve parameters E:y 2 =x 3 +ax+b, generate a temporary public-private key pair: Q A =d A G, Q B =d B G;

[0084] Shared key calculation K = d A Q B =d B Q A =(x K ,y K ), and finally extract the x coordinate as the symmetric key;

[0085] Step 2, data transmission encryption stage:

[0086] AES-GCM encryption (commonly used symmetric algorithm for SSL):

[0087] 2.4) Encrypt the data stream in blocks and generate authentication tags:

[0088] (Ciphertext, Tag) = AES_GCM_Encrypt(K, IV, Plaintext, AAD), where AAD is additional authentication data and Tag is used to verify data integrity.

[0089] Full-process mathematical correlation model:

[0090] Step 1: End-to-end encryption verification mechanism:

[0091] Storage side: Face data is encrypted to generate hash fingerprint, H enc =SHA256(SM4(CBC, P));

[0092] Transmission end: SSL encryption with MAC verification.

[0093] MAC=HMAC(K mac , Ciphertxt||Seq_num), the receiver verifies the consistency between MAC and decrypted Henc;

[0094] Step 2, key lifecycle management:

[0095] Adopting a hierarchical encryption mechanism: the master key Kmaster is stored in the HSM hardware module; the data encryption key

[0096] K data =PRF(K master , Label); the session key Ksession is dynamically generated through the SSL handshake.

[0097] Based on the above content, the data storage module implements facial data storage encryption through SM4, implements transmission channel protection through SSL protocol, and combines elliptic curve key exchange and hierarchical key management to form a complete security closed loop.

[0098] (3) data processing module 103;

[0099] 3.1) Quality Verification Algorithm: An advanced facial quality assessment algorithm is used to verify the quality of collected facial images. This algorithm comprehensively considers multiple image feature indicators, such as eye openness, facial posture, and occlusion. An image is considered qualified only when all image quality indicators reach a preset threshold.

[0100] The specific process is as follows:

[0101] In the first step, let the collected face image be I, and there are n feature indices to be evaluated, which are denoted as f1, f2, ..., f n , the corresponding preset thresholds are t1, t2, ..., t n ;

[0102] Step 2, feature index quantization function: For each feature index f i , there exists a quantization function Q i(I), used to convert the image I into a quantitative value of the feature index. For example, the quantization function Q1(I) of the degree of eye openness can be expressed as a certain statistical value of the pixels in the eye area calculated by an image processing algorithm;

[0103] Step 3: Quality judgment condition: The quality judgment condition of image I can be expressed as: Where ∧ represents a logical AND operation. When this condition is met, the quality of image I is considered to be qualified.

[0104] 3.2) Feature Extraction Algorithm: A facial feature extraction algorithm using deep learning technology is used to extract features from qualified facial images. The extracted feature values ​​are highly discriminative and stable, and can accurately represent the biometric characteristics of the face.

[0105] The specific process is as follows:

[0106] Step 1, data preprocessing: preprocess the qualified face image, including normalization, histogram equalization and other operations to improve the quality and consistency of the image. Let the preprocessed face image be I′, the deep learning model be M, and the model parameter be θ;

[0107] Step 2: Deep learning model training: Use a large amount of face image data to train the deep learning model. During the training process, the loss function L is usually used to measure the difference between the model output and the true label. For example, in the face recognition task, the commonly used loss function is triple loss, and its mathematical expression is: L triplet =max(||f a -f p || 2 -||f a -f n || 2 +α, 0), where fa is the feature vector of the anchor sample, fp is the feature vector of the positive sample (the sample belongs to the same category as the anchor sample), fn is the feature vector of the negative sample (the sample belongs to a different category than the anchor sample), and α is a positive boundary value. The training goal is to minimize the loss function Ltriplet by optimizing the model parameters θ.

[0108] In the third step, the preprocessed face image is input into the trained deep learning model, and the model outputs the feature vector of the face image. The output of the model M for the image I′ can be expressed as f=M(I′;θ), where f is the extracted face feature vector, which is usually a high-dimensional vector.

[0109] (4) data distribution module 104;

[0110] 4.1) Secure Transmission: SSL / TLS protocols are used to encrypt the transmission channel to ensure that data cannot be stolen or tampered with during transmission. When establishing a connection, the system will conduct strict identity authentication, and only authorized business systems can obtain facial copy data. At the same time, integrity verification is performed on the transmitted data to prevent data damage or loss during transmission.

[0111] The specific process is as follows:

[0112] Step 1: Establish an encrypted channel (SSL / TLS):

[0113] The client and server negotiate an encryption algorithm, verify the validity of the digital certificate, and generate a session key:

[0114] K session =KeyExchange(PK server , -SK client ), where PK server is the server public key, SK client is the client private key, K session is the negotiated symmetric key;

[0115] Step 2, Identity Verification:

[0116] The business system needs to provide a digital certificate or token, and the server verifies its legitimacy:

[0117] Verify(Sig token , PK issuer )=True, where Sig token Sign the token, PK issuer Public key of the issuing authority;

[0118] Step 3: Data encryption transmission:

[0119] Use symmetric encryption algorithm to encrypt face data to ensure confidentiality during transmission: C = E (K session , M), where C is the ciphertext, M is the plaintext data, and E is the encryption function;

[0120] Step 4, integrity check:

[0121] Generate a hash summary for the transmitted data and verify the integrity through signature or MAC: H(M) = SHA256(M), Verify(H(M), Sig hash )=True, where H(M) is the hash value, Sig hash The signature generated for the server.

[0122] 4.2) Implicit watermark embedding and traceability: Implicit watermarking is performed on facial data before it is sent to the facial business system, and real data traceability is provided to enable the source of facial data to be traced in the event of security incidents such as data leaks.

[0123] The specific process of implicit watermark embedding is as follows:

[0124] Step 1: Watermark information generation and encryption:

[0125] Traceability identification construction: Let the business system set be S = {s1, s2, ..., s m}, the distribution timestamp is t, and a unique watermark ID is generated w =H(K master ||s i ||t)mod 2 N , where H is the SHA-256 hash function, K maser is the master key, N is the number of watermark bits (usually 64-128 bits);

[0126] Watermark fragment embedding: Based on the improved LSB (least significant bit) algorithm or pseudo column watermark technology, the ID w Dispersed embedding of facial feature vectors, facial feature matrix (d is the feature dimension), the embedding rule is: where b k ∈{0, 1} is the k-th binary value of the watermark, and α is the scaling factor (usually 0.01-0.05) to ensure that the watermark is imperceptible;

[0127] Step 2: Data distribution and watermark association:

[0128] Mapping table storage: Maintain watermark-distribution record mapping table W map , the data structure is: W map (ID w )={s i ,t,IP address, operator ID}, are stored in blockchain or encrypted database to prevent tampering;

[0129] The specific process of implicit watermark tracing is as follows:

[0130] Step 1, watermark extraction and decoding:

[0131] Feature vector sampling: extract feature vector F from leaked data leak , and reverse calculation of watermark embedding rules: Where j is the predefined watermark embedding position index;

[0132] Identity reconstruction: Correct noise interference through majority voting algorithm and reconstruct the complete ID w :

[0133] Where δ is the Kronecker function, ensuring the fault tolerance rate ≤ 5%;

[0134] Step 2: Traceability verification:

[0135] Record query and matching: In W map Search Verify the source of the leak:

[0136] Matching conditions: Supports fuzzy queries to cope with timestamp deviation.

[0137] 4.3) Authorized Access Control: Establish a role-based access control (RBAC) mechanism to assign different access rights to different business systems. For example, the dormitory management system can only access facial data related to dormitory access authorization, and the library management system can only access facial data related to library borrowing authorization. This ensures that facial data is only used within the authorized scope.

[0138] The specific process is as follows:

[0139] Step 1: Role-authority matrix modeling:

[0140] Role definition: Construct role set R = {r1, r2, ..., r k}, such as: r1: dormitory management system role; r2: library management system role;

[0141] Permission definition: Create an operation permission set P = {p1, p2, ..., p m For example: p1: access dormitory access control facial data; p2: access library borrowing facial data;

[0142] Permission allocation relationship: Building a role-permission allocation matrix Example; PA(r1, p1) = 1, PA(r2, p2) = 1;

[0143] Step 2, user-role mapping:

[0144] User set: U = {u1, u2, ..., u n};

[0145] User role assignment: Create a user-role assignment matrix

[0146] Example: Dormitory management system u1 corresponds to UA(u1, r1) = 1:

[0147] Step 3: Data resource isolation:

[0148] Data classification: Define the face data set D = {d1, d2, ..., d l For example: d1: dormitory access control facial feature vector library; d2: library borrowing facial feature vector library;

[0149] Permission-resource binding: through permission p j Related data scope constraints: Example: Scope(p1) = {d1};

[0150] Step 4, access control verification:

[0151] When the business system i Request access to data k hour:

[0152]

[0153] Access is allowed only if the predicate is true.

[0154] (5) data destruction module 105;

[0155] Cross-system data destruction: When an individual withdraws authorization to use facial data, the system first destroys the user's facial data within this system. It then sends a destruction instruction to all business systems that have access to the user's facial data through a predefined interface. Upon receiving the instruction, the business system immediately deletes the relevant data stored locally and reports the destruction result to this system.

[0156] The specific process is as follows:

[0157] Step 1: Trigger conditions and event modeling:

[0158] User authorization revocation event: Assume that the user set U, when user u∈U triggers authorization revocation, a destruction event is generated: E(u) = {(u, t0)}, where t0 is the event triggering time;

[0159] Data status marking: The system maintains the user data status matrix D(u)∈{0,1}, where D(u)=1 indicates that the data exists and 0 indicates that it has been destroyed;

[0160] Step 2: Destruction of data in this system:

[0161] Logical deletion: mark the deleted record in the database: D(u)←0, operation time t1=t0+Δt1;

[0162] Physical Erase: Calling the secure erase algorithm to overwrite the storage media:

[0163] SecureErase(u)=Overwrite(Data(u), Pattern), where Pattern is a random overwrite sequence that complies with NIST standards;

[0164] Step 3: Cross-system command broadcast:

[0165] Interface definition: The predefined destruction instruction format is Msg = {uSig (u)}, where sig is a message authentication code based on hash: Sig (u) = HMAC (K shared ,u||t1);

[0166] Broadcast mechanism: to all associated business system sets S = {s1, s2, ..., s n Send command:

[0167] Broadcast(si, Msg):

[0168] Step 4: Business system response and feedback:

[0169] Atomic operations: business systems i Execute after receiving the instruction:

[0170]

[0171] Feedback verification: return response Where t2 is the time it takes to complete the operation;

[0172] Step 5, status synchronization and auditing:

[0173] Aggregation result: The main system verifies the response signatures of all business systems and updates the global status:

[0174] Only when all Delete(s i , u) = 1, it is determined that the cross-system destruction is successful;

[0175] Logging: Write to the audit log L to meet compliance requirements:

[0176]

[0177] (VI) Supervision Analysis Module 106

[0178] 6.1) Supervision log records;

[0179] The system keeps detailed records of the entire life cycle of facial data, including data collection time, collection personnel, authorized business system, destruction time, and other information. This information is stored in the database in the form of logs to provide data support for subsequent regulatory analysis.

[0180] The specific process is as follows:

[0181] Step 1: Full life cycle event modeling:

[0182] Event type definition: Construct operation set E = {e1, e2, ..., e k}, including: e1: data collection (time, operator, device ID); e2: data storage (storage location, encryption status); e3: data access (business system ID, access purpose); e4: data destruction (trigger condition, execution result);

[0183] Log entry generation: Each event generates a log record L i =(t i , o i ,u i , d i , s i ), where: i : Timestamp (accurate to milliseconds); o i :Operation type (such as O i ∈E);u i :Operation subject (personnel ID / system ID); d i : Data identification (unique hash value of face data); s i : Operation status (success / failure / exception code);

[0184] Step 2: Log storage mechanism: Use a time series database to store logs, ensuring that each record meets the requirements; use hash chain technology to ensure that logs cannot be tampered with; support log retrieval by time range, statistics by operation frequency, and anomaly detection;

[0185] Among them, anomaly detection is based on a rule engine, defining abnormal conditions such as:

[0186] where Δt min : Preset the minimum operation interval, such as multiple sensitive operations within 1 second.

[0187] 6.2) Real-time monitoring

[0188] Supervisors can view the use of facial data in real time through the system interface, such as which business systems use facial data, and on this basis, abnormal access detection is also possible;

[0189] The abnormal access detection process is as follows:

[0190] Step 1, anomaly detection model:

[0191] Access frequency analysis based on Poisson distribution: λ=historical average access rate, when P(X≥k current)<0.01, an alarm is triggered;

[0192] Step 2, data thermal model:

[0193] Calculation of facial data access heat: α and β are adjustment coefficients used to balance time decay and safety level;

[0194] Step 3, real-time event stream processing:

[0195] Sliding window statistics (such as 5-minute window):

[0196] W is all events in the current time window;

[0197] Application example, abnormal access detection in library system:

[0198] When the library system S lib More than 100 facial data read requests are initiated within 10 minutes (historical average λ = 20): Triggering a level 1 alarm;

[0199] The monitoring interface automatically highlights the node in red and displays the dormitory access control data D that it accesses. dorm , prompting the risk of exceeding authority.

[0200] 6.3) Data analysis and report generation

[0201] The system regularly analyzes regulatory logs and generates various statistical reports, such as facial data usage trend reports and data leakage risk assessment reports. These reports can help university management understand the usage of facial data, promptly identify potential security risks, and take corresponding measures to prevent them.

[0202] According to the above system composition: this facial data full life cycle management system proposes the concept of full life cycle management of facial data of universities for the first time, covering all aspects from data collection to destruction, and realizing comprehensive supervision of facial data; it has the unification of multiple algorithm standards and can adapt to various brands of facial recognition quality algorithms, such as Megvii, SenseTime, Baidu, etc., solving the problem of compatibility of quality standards between different algorithms.

[0203] like Figure 2 As shown in FIG, a method for managing the entire life cycle of facial data is provided, and the specific steps are as follows:

[0204] Step 1: Personal Authorization: Before facial data collection, the system displays an authorization agreement to the individual, clearly informing them of the purpose of collection, scope of use, storage period, and other information about their personal data. The individual then authorizes the operation through the system interface.

[0205] Step 2: Face collection and verification: Facial images are collected using mobile phone cameras according to unified standards and procedures. The collected facial images are then verified using a quality assessment algorithm to ensure they are clear, unobstructed, and free of blur. At the same time, the collected facial images are linked to personal identification information for public security facial identity verification.

[0206] Step 3: Secure storage and distribution: The verified facial data is encrypted using a national encryption algorithm and stored in the database. When other business systems need to use facial data, the system distributes the matching data to the relevant business systems via SSL secure encryption.

[0207] Step 4: Authorized access control: Establish a role-based access control mechanism to assign different access rights to different business systems, ensuring that facial data is only used within the authorized scope;

[0208] Step 5: Full life cycle supervision analysis: The system conducts full life cycle supervision analysis on the user's facial data from generation to destruction, records information such as data collection time, verification status, usage, destruction time, and generates a full supervision log; supervisors can view the supervision log through the system interface and conduct real-time monitoring and analysis of the use of facial data.

[0209] According to the above process: comprehensive management of personal authorization of facial data, facial collection, identity and quality verification, national secret storage and distribution, revocation and destruction, traceability and other links, to achieve full life cycle supervision of facial data of college teachers and students from generation to destruction, and ensure the security of facial data and personal information.

[0210] like Figure 3 As shown, in step 3, add the face tracing function. The specific steps are as follows:

[0211] Step 3.1: Watermark information generation and encryption: Construct the traceability mark and embed the watermark fragment;

[0212] Step 3.2: Data distribution and watermark association: Construct a watermark-distribution record mapping table and store it in encrypted form;

[0213] Step 3.3: When data leakage occurs, perform implicit watermark tracing;

[0214] Step 3.4: Watermark extraction and decoding: feature vector sampling and logo reconstruction;

[0215] Step 3.5: Provenance Verification: Record query and matching, and perform the query in the watermark-distribution record mapping table based on the reconstruction identifier.

[0216] According to the above process: an implicit digital watermark is stamped on the issued negative film; when a security incident such as data leakage occurs, the system can trace the source of the data through the digital watermark information and determine the responsible party.

[0217] like Figure 4 As shown, in step 5, the face data is revoked and destroyed. The specific steps are as follows:

[0218] Step 5.1: Trigger conditions and event modeling: User triggers the authorization withdrawal event and data status mark;

[0219] Step 5.2: Destroy the data in this system: perform logical deletion and physical removal;

[0220] Step 5.3: Cross-system command broadcast: pre-define the destruction command and use the broadcast mechanism to send the command to all related business systems;

[0221] Step 5.4: Business system response and feedback;

[0222] Step 5.5: State synchronization and auditing: The main system verifies the response signatures of all business systems, updates the global state, and writes to the audit log.

[0223] According to the above process: when an individual withdraws the authorization to use facial data, the system uses cross-system data destruction technology to quickly and uniformly destroy the relevant data from this system to all business systems that obtain the user's facial data; the destruction process follows strict data destruction policies to ensure that the data cannot be recovered during the destruction process.

[0224] The foregoing is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained herein shall, unless otherwise specified or limited, be implemented in accordance with conventional means in the art.

Claims

1. A facial data full life cycle management system, comprising a full life cycle management system (1), characterized in that: The full life cycle management system (1) is internally provided with a data acquisition module (101), a data storage module (102), a data processing module (103), a data distribution module (104), a data destruction module (105) and a supervision and analysis module (106); The data storage module (102) uses the national secret encryption algorithm SM4 to encrypt and store the data in the face database; during the data transmission process, the SSL protocol is used to encrypt the data; The data distribution module (104) uses the SSL / TLS protocol to encrypt the transmission channel; when establishing a connection, the system will perform strict identity authentication, and only authorized business systems can obtain face copy data; at the same time, integrity verification is performed on the transmitted data; The data distribution module (104) uses implicit watermark embedding and tracing to perform implicit watermark processing when sending to the face business system, and provides real data tracing capability for tracing the source of face data in the event of security incidents such as data leakage; The data distribution module (104) adopts an authorized access control mechanism: establishing a role-based access control mechanism to assign different access rights to different business systems, ensuring that facial data is only used within the authorized scope; The data destruction module (105) adopts cross-system data destruction: when an individual withdraws the authorization to use facial data, the system first destroys the facial data of the user in the system; then, a destruction instruction is sent to all business systems that have obtained the facial data of the user through a predefined interface; after receiving the instruction, the business system immediately deletes the relevant data stored locally and feeds back the destruction result to the system.

2. The facial data full lifecycle management system according to claim 1, characterized in that: The data acquisition module (101) is used for the system to perform quality assessment on the acquired images in real time during the acquisition process of the mobile phone camera.

3. The facial data full lifecycle management system according to claim 1, characterized in that: The data processing module (103) adopts a quality verification algorithm to perform quality verification on the collected facial image. The algorithm comprehensively considers the characteristic indicators of the image. Only when the various image quality indicators reach a preset threshold, the image quality is considered to be qualified.

4. The facial data full lifecycle management system according to claim 1, characterized in that: The data processing module (103) adopts a feature extraction algorithm to extract features from a human face image of qualified quality. The extracted feature values ​​have high discrimination and stability and are used to accurately represent the biological features of the human face.

5. The facial data full lifecycle management system according to claim 1, characterized in that: The supervision analysis module (106) includes a supervision log recording unit, and the system records the entire life cycle of facial data in detail, including information such as the data collection time, collection personnel, authorized business system, and destruction time.

6. The facial data full lifecycle management system according to claim 1, characterized in that: The supervision and analysis module (106) includes a real-time monitoring unit, and supervisors can view the use of facial data in real time through the system interface, and also has abnormal access detection.

7. The facial data full lifecycle management system according to claim 1, characterized in that: The supervision analysis module (106) includes a data analysis and report generation unit. The system regularly analyzes the supervision log and generates various statistical reports.

8. A method for managing the entire life cycle of facial data, using the facial data entire life cycle management system according to claim 7, characterized in that: The specific steps are as follows: Step 1: Personal Authorization: Before facial data collection, the system displays an authorization agreement to the individual, clearly informing them of the purpose of collection, scope of use, storage period, and other information about their personal data. The individual then authorizes the operation through the system interface. Step 2: Face collection and verification: Facial images are collected using mobile phone cameras according to unified standards and procedures. The collected facial images are then verified using a quality assessment algorithm to ensure they are clear, unobstructed, and free of blur. At the same time, the collected facial images are linked to personal identification information for public security facial identity verification. Step 3: Secure storage and distribution: The verified facial data is encrypted using a national encryption algorithm and stored in the database. When other business systems need to use facial data, the system distributes the matching data to the relevant business systems via SSL secure encryption. Step 4: Authorized access control: Establish a role-based access control mechanism to assign different access rights to different business systems, ensuring that facial data is only used within the authorized scope; Step 5: Full life cycle supervision analysis: The system conducts full life cycle supervision analysis on the user's facial data from generation to destruction, records information such as data collection time, verification status, usage, destruction time, and generates a full supervision log; supervisors view the supervision log through the system interface and conduct real-time monitoring and analysis of the use of facial data.

9. A facial data full lifecycle management system and method according to claim 8, characterized in that: In step 3, add the face tracing function. The specific steps are as follows: Step 3.1: Watermark information generation and encryption: Construct the traceability mark and embed the watermark fragment; Step 3.2: Data distribution and watermark association: Construct a watermark-distribution record mapping table and store it in encrypted form; Step 3.3: When data leakage occurs, perform implicit watermark tracing; Step 3.4: Watermark extraction and decoding: feature vector sampling and logo reconstruction; Step 3.5: Provenance Verification: Record query and matching, and perform the query in the watermark-distribution record mapping table based on the reconstruction identifier.

10. A facial data full lifecycle management system and method according to claim 8, characterized in that: In step 5, the facial data is revoked and destroyed. The specific steps are as follows: Step 5.1: Trigger conditions and event modeling: User triggers the authorization withdrawal event and data status mark; Step 5.2: Destroy the data in this system: perform logical deletion and physical removal; Step 5.3: Cross-system command broadcast: pre-define the destruction command and use the broadcast mechanism to send the command to all related business systems; Step 5.4: Business system response and feedback; Step 5.5: State synchronization and auditing: The main system verifies the response signatures of all business systems, updates the global state, and writes to the audit log.

Images