Iris template multi-level life cycle management and dynamic drift elimination algorithm

Abstract

The invention discloses a multi-level life cycle management and dynamic drift elimination algorithm for an iris template, and relates to the technical field of biological feature recognition and information security. A three-layer hierarchical storage architecture of a module layer, an edge layer and a cloud layer is constructed, and a dual-index decision-making mechanism of drift accumulation monitoring and activeness assessment is combined, so that the multi-level life cycle management and dynamic drift elimination algorithm for the iris template is realized. The intelligent life cycle management of the iris template is realized, a differential chain synchronization technology and a hierarchical key pool encryption system are adopted, the access performance and the storage cost are optimized while the data security is guaranteed, and a dynamic elimination algorithm processes the life cycle of the iris template through a self-destruction countdown mechanism according to the user activeness difference, so that the safety of the iris template is improved. The method effectively solves the limitation of a traditional fixed expiration strategy, achieves the remarkable improvement in the aspects of recognition accuracy, response speed and safety protection, and is particularly suitable for application scenes with extremely high requirements for real-time performance and safety.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biometric feature recognition and information security, and in particular to an iris template multi-level life cycle management and dynamic drift elimination algorithm. BACKGROUND

[0002] The current iris recognition iris template multi-level life cycle management system generally adopts a fixed life cycle template management strategy, usually setting a uniform expiration time such as 30 days or 90 days for template cleaning. This rigid management method has significant technical defects: first, the fixed expiration strategy cannot reflect the actual usage pattern of the user, resulting in the deletion of effective templates of high-frequency users, while the degraded templates of low-frequency users are long-term retained in the iris template multi-level life cycle management system; second, the traditional iris template multi-level life cycle management system lacks real-time monitoring capability for template quality degradation, and cannot timely discover and handle the template drift problem caused by factors such as physiological changes and device aging; third, the single-layer storage architecture has performance bottlenecks when facing large-scale deployment, making it difficult to balance access speed and storage cost; finally, the template deletion operation of the existing iris template multi-level life cycle management system is usually irreversible, lacking the necessary audit traceability capability.

[0003] These technical limitations are particularly prominent in application scenarios with extremely high security and real-time requirements. Users frequently encounter recognition failures during high-frequency use, not only affecting user experience, but also potentially causing security risks at critical moments, while the long-term existence of degraded templates increases the risk of attacks on the iris template multi-level life cycle management system, and the lack of effective template quality evaluation mechanism makes the iris template multi-level life cycle management system unable to actively prevent potential security threats. SUMMARY

[0004] The technical problem solved by the present application is to provide an iris template multi-level life cycle management and dynamic drift elimination algorithm, which places template security protection and recognition accuracy optimization in a unified technical framework, realizes their collaborative optimization, uses a three-layer hierarchical storage architecture to handle the spatial distribution of templates, and uses drift accumulation monitoring and activity evaluation algorithms to solve the temporal evolution management problem of templates, thereby maximizing recognition accuracy and user experience while ensuring the security of the iris template multi-level life cycle management system.

[0005] To solve the above technical problems, the present application provides the following technical solutions: The iris template multi-level life cycle management and dynamic drift elimination algorithm includes a module layer, an edge layer, a cloud layer and a dynamic management module. The module layer is used to store first cache data, process real-time acquisition templates and perform comparison operations, and the first cache data is encrypted by a session key. The edge layer is used for storing second cache data and processing rolling incremental messages, and the second cache data is encrypted by a daily key; The cloud layer is used for storing third cache data and archiving, and the third cache data is encrypted by a parent key; The dynamic management module is used for scheduling the iris template among the module layer, the edge layer and the cloud layer according to the drift accumulation of the iris template and the user activity, and performing a dynamic elimination operation.

[0006] Preferably, the comparison operation includes acquiring a current iris image of a user by an iris acquisition device, extracting iris features of the current iris image of the user using a Gabor filter or a deep learning algorithm, searching for a cached template of the user in a static random access memory, calculating a Hamming distance or an Euclidean distance between the iris features and the cached template, comparing the distance value with a preset threshold value, and if the distance value is less than the preset threshold value, matching is successful, and comparison data of the iris template is acquired, the comparison data including a comparison timestamp, a current comparison score and a registration baseline score. The module layer uses a static random access memory as a storage medium, the capacity of the static random access memory is within a first threshold range, the access delay time is within a first time range, and the static random access memory is invalid when powered off; The edge layer uses a non-volatile memory express solid state disk as a storage medium, the capacity of the non-volatile memory express solid state disk is within a second threshold range, the access delay time is within a second time range, and the non-volatile memory express solid state disk is used for storing daily incremental data; The cloud layer uses a hardware security module to encrypt object storage, the capacity of the hardware security module encrypted object storage is PB level, the access delay time is within a third time range, and the hardware security module encrypted object storage is used for blockchain-based audit backtracking; The first cache data includes cache data within a first unit of time; The second cache data includes cache data within a second unit of time; The third cache data includes cache data within a third unit of time; The blockchain audit backtracking includes integrity verification of packaged incremental differential data by a Merkle tree, and a root hash of the Merkle tree is recorded in the blockchain, when verifying the integrity of the incremental differential data, the Merkle root is reconstructed and compared with the root hash recorded in the blockchain, and if they are consistent, the verification is passed.

[0007] Preferably, the module layer, the edge layer and the cloud layer constitute a three-layer hierarchical storage architecture; A hierarchical key pool encryption system is used among the three-layer hierarchical storage architecture; The hierarchical key pool encryption system includes a session key, a day key, and a master key; The session key, day key, and master key are used for data protection and access control, respectively, specifically including: The module layer encrypts the iris template using a session key for a single session, and the invalidation status of the session key is determined based on the session end status. The session key becomes invalid when the session ends, and remains valid when the session continues. The edge layer encrypts the incremental differential data using a daily rotating key; The cloud layer encrypts the iris template using an annually rotating master key, and the master key is protected by a hardware security module. The decryption process verifies the master key, day key, and session key in sequence. The session key has a lifespan of one session, and AES-256-GCM is used as the first encryption algorithm. The lifespan of the daily key is 24 hours, and AES-256-CBC is used as the second encryption algorithm; The lifecycle of the master key is rotated annually, and AES-256-XTS is used as the third encryption algorithm.

[0008] Preferably, the dynamic management module includes: When the iris template in the module layer has no access request within the fourth unit of time, the iris template in the module layer is migrated to the edge layer and the local cache data of the module layer is deleted. When the user activity level of the iris template in the edge layer is less than the first preset threshold within the fifth unit of time, the iris template in the edge layer will be migrated to the cloud layer for long-term archiving. When the iris template in the cloud layer has no access records within the sixth unit of time, the encrypted destruction process is triggered after the administrator's review, and the relevant records of the iris template destruction are recorded in the distributed ledger. The relevant records include the destruction time, the operating entity, and the associated template identification information.

[0009] Preferably, the module layer and the edge layer, and the edge layer and the cloud layer, transmit data via a differential chain synchronization mechanism, which includes: The formula for calculating the incremental difference data is: Δk=T k ⊕T k-1 ; Where Δk represents the incremental difference data between the iris template data collected in the k-th acquisition and the iris template data collected in the (k-1)-th acquisition, and T k T represents the iris template data collected in the kth iteration. k-1represents the iris template data collected at the k-1th time.

[0010] Preferably, the user activity is calculated by an exponential decay model on the number of successful comparisons in the seventh unit time, and the formula of the user activity is: A =∑e^(-(t0-t i ) / τ); wherein A is the user activity, t0 is the current time, t i is the i th successful comparison timestamp, and τ is a time decay constant. The drift accumulation is taken as a quantitative index of iris template quality decay, and the feature distance deviation of each comparison is accumulated by a sliding window, and the calculation formula of the feature distance deviation is: Feature distance deviation =∑|current comparison score - registration baseline score|; The feature distance includes a Hamming feature distance, a Gabor filter feature distance, or a deep learning feature distance. When the drift accumulation reaches a second preset threshold, the iris template state is evaluated in combination with the user activity and a risk factor. The risk factor includes a multi-dimensional security index of abnormal login attempts, failed liveness detection, and abnormal geographic location.

[0011] Preferably, the dynamic elimination operation includes: If the elimination condition is met, the iris template multi-level life cycle management system starts a self-destruction countdown mechanism. If successful liveness verification and iris template matching are detected during the self-destruction countdown, the self-destruction countdown mechanism is cancelled and the drift accumulation is reset. The calculation formula of the length of the self-destruction countdown mechanism is: t wipe = λ × e^(-A); wherein t wipe represents the length of the self-destruction countdown, λ is a strategy parameter, and A is the user activity. The hierarchical management strategy of the drift accumulation is: When the drift accumulation is less than or equal to a third preset threshold, the iris template quality is good, the iris template multi-level life cycle management system operates normally, and continues to monitor without intervention. When the drift accumulation is greater than the third preset threshold but less than or equal to a second preset threshold, the iris template has obvious drift, the monitoring frequency is increased, and the iris template is updated. When the drift accumulation is greater than the second preset threshold, the iris template quality is seriously deteriorated, and the self-destruction countdown mechanism is started.

[0012] Preferably, the user activity and the countdown length are negatively correlated by the countdown length calculation formula; The evaluation period of the risk factor and the sliding window period of the drift accumulation amount are consistent, and when the risk factor is greater than a first threshold value and the user activity is less than a second threshold value, a second preset threshold value is automatically reduced, accelerating and accelerating the iris template elimination process; If a successful liveness verification and iris template matching are detected during the self-destruction countdown period, the iris template multi-level life cycle management system cancels the countdown and resets the drift accumulation amount; The liveness verification includes verifying whether the real-time collected iris template information is from a real human body through a biometric anti-counterfeiting technology; The iris template matching includes comparing the features of the real-time collected and processed iris template with the reference iris template, calculating the feature distance deviation between them, and determining whether they are the iris templates of the same user. When the feature distance deviation reaches a second preset threshold value, it is determined that the matching is successful, indicating that it is the iris template of the same user, and the identity authentication is completed. When the feature distance deviation does not reach the second preset threshold value, it is determined that the matching fails; The reference iris template includes an initial iris template determined by the user during the registration stage of the iris template multi-level life cycle management system.

[0013] Preferably, the module layer, the edge layer and the cloud layer achieve hierarchical security protection through physical isolation; The module layer is used as a cache layer; The edge layer is used as an intermediate cache layer; The cloud layer is used as a permanent storage layer; The overall comparison delay time of the three-layer hierarchical storage architecture is controlled within an eighth unit of time.

[0014] The iris template dynamic drift elimination algorithm is applied to the iris template multi-level life cycle management system and includes the following steps: Step S100: Real-time acquisition of iris template comparison data; Step S200: Calculate the drift accumulation amount by accumulating the feature distance deviation, and calculate the user activity by an exponential decay model; Step S300: Combine the risk factor to evaluate the iris template state; Step S400: Perform hierarchical management strategy according to the drift accumulation amount; Step S500: Perform dynamic elimination operation according to the hierarchical management strategy.

[0015] The beneficial effects of the present application: through technical architecture innovation and algorithm optimization, significant improvement is achieved in multiple key indicators, the overall comparison delay control of the iris template multi-level life cycle management system is within 25ms, compared with the traditional cloud centralized processing method, the response speed is improved by more than 60%; the differential chain synchronization mechanism reduces the uplink bandwidth demand by more than 75%, greatly reducing the network transmission cost and delay risk; the dynamic threshold adjustment algorithm reduces the false alarm rate by 15% and the rejection rate by 12%, significantly improving the recognition accuracy and user experience of the iris template multi-level life cycle management system; in terms of security, the hierarchical key pool encryption system and the design of power failure invalidation greatly improve the anti-leakage capability of the iris template multi-level life cycle management system, even if one level is attacked, it will not affect the security of the overall iris template multi-level life cycle management system; the differential chain synchronization mechanism and the audit backtracking of the block chain ensure the traceability and non-repudiation of the operation, meet the compliance requirements of the level protection and data security law. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The basic flowchart of the iris template multi-level life cycle management and dynamic drift elimination algorithm provided by an embodiment of the present application is shown in the figure. Figure 2 The step flowchart of the iris template multi-level life cycle management and dynamic drift elimination algorithm provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present application, not all embodiments.

[0018] Embodiment one, identity verification of public security police station The iris template multi-level life cycle management system is deployed in the household registration window of a city public security bureau police station. The user iris collection and comparison frequency is high, and the real-time requirement is strict. The iris template multi-level life cycle management system adopts a three-layer hierarchical storage architecture for deployment. The window terminal is equipped with a static random access memory cache as a module layer. An edge gateway is deployed in the police station to undertake the function of the middle layer. The iris template multi-level life cycle management system is connected to the city bureau cloud hardware security module through a private network. In the initial deployment stage, the iris template multi-level life cycle management system registers the iris of each person who comes to the window for business. The complete template T0 is stored in the cloud hardware security module and is encrypted and protected by using an AES-256-GCM algorithm and a parent key. When the people come to the window for business again, the iris template multi-level life cycle management system first searches the cache template in the static random access memory of the module layer. If there is no hit, the edge gateway is requested. The edge gateway maintains the templates of active users for nearly 30 days and stores them by using a daily key. For a first-time access user, the iris template multi-level life cycle management system downloads the baseline template and the related incremental chain from the cloud and reconstructs the current effective template in the edge layer. In actual operation, the iris template multi-level life cycle management system continuously monitors the drift of each comparison. Taking Mr. Zhang as an example, the baseline template of Mr. Zhang gradually drifts during the continuous use. The iris template multi-level life cycle management system records that the cumulative drift reaches 0.28 on the 15th day, close to the preset threshold 0.32. Since Mr. Zhang is a high-frequency user, the activity calculation value is 8.5. The iris template multi-level life cycle management system determines that the template is still in the acceptable range. However, on the 22nd day, the cumulative drift breaks through the threshold. The iris template multi-level life cycle management system starts the elimination evaluation. Considering the high activity of Mr. Zhang, the countdown is set to t wipe = 60 x e^(-8.5) ≈ 0.12 seconds, which is extremely short. During the actual countdown process, Mr. Zhang successfully completes a live body verification and template matching. The iris template multi-level life cycle management system immediately cancels the countdown, resets the drift cumulative value, and generates a new incremental Δ 22 update template. After six months of actual operation, the average handling time of the window of the police station is shortened from 38 seconds to 29 seconds, and the user satisfaction is significantly improved. The false positive rate of the iris template multi-level life cycle management system is controlled below 0.01%, and the rejection rate is reduced to 0.08%, reaching the expected performance indicators.

[0019] Embodiment two, mobile police terminal offline identification A provincial public security bureau equipped mobile police tablet for front-line police officers needs to perform identity verification in a network-limited environment, which requires extremely high offline processing capability. The multi-level iris template lifecycle management system adopts an edge enhancement deployment mode, integrates a larger capacity of non-volatile memory fast solid state disk storage on the tablet device, and can cache the iris templates of high-frequency target personnel in the region. The multi-level iris template lifecycle management system realizes an efficient template synchronization mechanism through a differential chain synchronization mechanism. The tablet device downloads the template benchmark version and incremental chain data of the target region from the cloud regularly in a networked state. Due to the use of XOR differential encoding, the data volume of each incremental data is usually only 3 to 6 KB, which greatly reduces the data transmission volume. The multi-level iris template lifecycle management system generates a Merkle tree for the collected incremental data every 5 minutes, uploads the root hash to the blockchain network, and ensures the verifiability of data integrity. During the actual execution of the task, police officer Li needs to perform identity verification on a suspicious person during a night patrol. Due to the weak network signal at the time, the multi-level iris template lifecycle management system switches to offline mode, and the iris template of the suspicious person Wang is stored in the edge layer of the tablet. The multi-level iris template lifecycle management system completes the template reconstruction and comparison process locally, and the entire recognition process takes less than 40 milliseconds. During the comparison process, the multi-level iris template lifecycle management system detects a slight drift (AH=0.15), but since Wang is marked as a medium activity user (A=3.2) in the multi-level iris template lifecycle management system, the drift accumulation has not yet reached the threshold, and the multi-level iris template lifecycle management system normally completes the identity confirmation. The mobile multi-level iris template lifecycle management system has accumulated more than 12,000 identity verification requests in three months of pilot operation, with an offline recognition success rate of 98.5%, an average response time of less than 35 milliseconds, and an effective improvement in the law enforcement efficiency and security protection capability of front-line police officers.

[0020] Example Three: High-security level identity authentication for financial institutions An important business place of a state-owned bank needs to strictly authenticate the identity of customers, and the multi-level life cycle management system of the iris template faces high concurrent access and extremely strict security requirements. The multi-level life cycle management system of the iris template adopts a security-enhanced configuration, and all three levels are deployed with additional security measures. The layered key pool of the multi-level life cycle management system of the iris template adopts a more complex double-encapsulation structure, the session key is updated immediately after each interaction, the daily key uses a true random number generated by a hardware security module, the master key is periodically rotated and protected by a multi-signature mechanism, and the iris template of each customer is protected by three layers of encryption. Even if the key of a certain level is leaked, an attacker cannot obtain complete template information. In the actual business process, a large customer, Zhao, frequently handles business, and the multi-level life cycle management system of the iris template maintains a complete use history for him. Through the activity algorithm analysis, the A value of Zhao remains high (A>10) for a long time, and the risk factor is always low (R<0.1). When the multi-level life cycle management system of the iris template detects that the template of Zhao has a slight drift, it does not immediately start the elimination process, but optimizes the template quality through the incremental update mechanism. In contrast, a low-frequency customer, Qian, has only handled two businesses in six months. After the cumulative template drift of his template reached the threshold, the multi-level life cycle management system of the iris template quickly completed the template elimination, avoiding potential security risks. The multi-level life cycle management system of the iris template of the bank has been running for one year, and has handled more than 500,000 customer identity authentication processes, with zero false recognition rate and zero security incidents, and has received high recognition from the regulatory authorities.

[0021] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. An iris template multi-level life cycle management and dynamic drift eviction algorithm, characterized by, The system comprises a module layer, an edge layer, a cloud layer and a dynamic management module. The module layer is used for storing first cache data, processing real-time collection templates and performing comparison operations, the first cache data being encrypted by a session key; The edge layer is used for storing second cache data and processing rolling incremental messages, the second cache data being encrypted by a daily key; The cloud layer is used for storing third cache data and archiving, the third cache data being encrypted by a parent key; The dynamic management module is used for scheduling iris templates among the module layer, the edge layer and the cloud layer according to the drift accumulation of the iris templates and the user activity, and performing dynamic elimination operations.

2. The iris template multi-level life cycle management and dynamic drift elimination algorithm of claim 1, wherein the comparison operation comprises obtaining a current iris image of a user by an iris collection device, extracting iris features of the current iris image of the user by using a Gabor filter or a deep learning algorithm, searching for a cached template of the user in a static random access memory, calculating a Hamming distance or an Euclidean distance between the iris features and the cached template, comparing the distance value with a preset threshold, and if the distance value is less than the preset threshold, matching is successful, and comparison data of the iris template is obtained, the comparison data comprising a comparison timestamp, a current comparison score and a registration baseline score. The module layer uses a static random access memory as a storage medium, the capacity of the static random access memory is within a first threshold range, the access delay time is within a first time range, and the static random access memory is invalid when powered off. The edge layer uses a non-volatile memory express solid state disk as a storage medium, the capacity of the non-volatile memory express solid state disk is within a second threshold range, the access delay time is within a second time range, and the non-volatile memory express solid state disk is used for storing daily incremental data. The cloud layer uses a hardware security module to encrypt object storage, the capacity of the hardware security module is PB level, the access delay time is within a third time range, and the hardware security module is used for blockchain-based audit backtracking. The first cache data comprises cache data within a first unit of time. The second cache data comprises cache data within a second unit of time. The third cache data comprises cache data within a third unit of time. The blockchain audit backtracking comprises integrity verification of packaged incremental differential data by a Merkle tree, the root hash of the Merkle tree being recorded in the blockchain, and when verifying the integrity of the incremental differential data, the Merkle root is reconstructed and compared with the root hash recorded in the blockchain, and if they are consistent, the verification is passed. The module layer, the edge layer and the cloud layer constitute a three-layer hierarchical storage architecture.

3. The iris template multi-level life cycle management and dynamic drift eviction algorithm of claim 2, wherein, The three-layer hierarchical storage architecture adopts a hierarchical key pool encryption system. The hierarchical key pool encryption system comprises a session key, a daily key and a parent key. The session key, the daily key and the parent key are respectively used for data protection and access control, specifically comprising: The module layer encrypts the iris templates by a session key of a single session, and the invalid state of the session key is determined according to the end state of the session. ​ The session key is valid when the session is not ended, and the session key is invalid when the session is ended; The edge layer encrypts the incremental differential data by daily rotating daily keys; The cloud layer encrypts the iris template by annually rotating master keys, and the master keys are protected by a hardware security module; The decryption process is verified in order of the master key, the daily key, and the session key; The session key has a single-session life cycle, and AES-256-GCM is used as the first encryption algorithm; The daily key has a 24-hour life cycle, and AES-256-CBC is used as the second encryption algorithm; The master key has an annual rotation life cycle, and AES-256-XTS is used as the third encryption algorithm.

4. The iris template multi-level life cycle management and dynamic drift eviction algorithm of claim 1, wherein, The dynamic management module includes: When there is no access request to the iris template in the module layer within a fourth unit of time, the iris template in the module layer is migrated to the edge layer, and the local cache data of the module layer is deleted; When the user activity of the iris template in the edge layer within a fifth unit of time is less than a first preset threshold, the iris template in the edge layer is migrated to the cloud layer for long-term archiving; When there is no access record to the iris template in the cloud layer within a sixth unit of time, an encryption destruction process is triggered after administrator review, and related records of the destruction of the iris template are recorded in a distributed ledger, including destruction time, operation subject, and associated template identification information.

5. The iris template multi-level life cycle management and dynamic drift eviction algorithm of claim 2, wherein, Transmission is performed between the module layer and the edge layer, and between the edge layer and the cloud layer through a differential chain synchronization mechanism, which includes: The calculation formula of the incremental differential data is: Δk = T k ⊕T k-1 ; wherein Δk represents incremental difference data between the iris template data of the kth acquisition and the iris template data of the (k-1)th acquisition, T k represents the iris template data of the kth acquisition, T k-1 represents the iris template data of the (k-1)th acquisition.

6. The iris template multi-level life cycle management and dynamic drift eviction algorithm of claim 1, wherein, The user activity is calculated by an exponential decay model on the number of successful comparisons within a seventh unit of time, and the formula of the user activity is: A =∑e^(-(t0-t i ) / τ); wherein A is the user activity, t0is the current time, t i is the ith successfully compared timestamp, and τ is a time decay constant. The drift accumulation is used as a quantitative indicator of the quality decay of the iris template, and the feature distance deviation of each comparison is accumulated through a sliding window, and the calculation formula of the feature distance deviation is: Feature distance deviation = Σ | current comparison score - registration baseline score |; The feature distance includes Hamming feature distance, Gabor filter feature distance, or deep learning feature distance; When the drift accumulation reaches a second preset threshold, the iris template state is evaluated in combination with the user activity and the risk factor; The risk factor includes multi-dimensional security indicators such as abnormal login attempts, liveness detection failures, and abnormal geographic locations.

7. The iris template multi-level life cycle management and dynamic drift eviction algorithm of claim 6, wherein, The dynamic elimination operation includes: If the elimination condition is met, the iris template multi-level life cycle management system starts a self-destruction countdown mechanism; If successful liveness verification and iris template matching are detected during the self-destruction countdown, the self-destruction countdown mechanism is canceled and the drift accumulation is reset; The calculation formula of the countdown length of the self-destruction countdown mechanism is: t wipe = λ x e^(-A); where t wipe denotes the length of the countdown to self-destruction, λ is a policy parameter, and A is the user activity. The hierarchical management strategy of the drift accumulation is: When the drift accumulation is less than or equal to a third preset threshold, the quality of the iris template is good, the iris template multi-level life cycle management system operates normally, and continues to monitor without intervention. When the drift accumulation is greater than the third preset threshold but less than or equal to the second preset threshold, the iris template has obvious drift, the monitoring frequency is increased, and the iris template is updated; When the drift accumulation is greater than the second preset threshold, the quality of the iris template is seriously degraded, and a self-destruction countdown mechanism is started.

8. The iris template multi-level life cycle management and dynamic drift eviction algorithm of claim 7, wherein, The negative correlation mapping between the user activity and the countdown length is performed through the countdown length calculation formula; The evaluation period of the risk factor is consistent with the sliding window period of the drift accumulation, when the risk factor is greater than the first threshold and the user activity is less than the second threshold, the second preset threshold is automatically reduced, and the elimination process of the iris template is accelerated; If successful liveness verification and iris template matching are detected during the self-destruction countdown, the iris template multi-level life cycle management system cancels the countdown and resets the drift accumulation; The liveness verification includes verifying whether the real-time collected iris template information is from a real human body through a biometric anti-counterfeiting technology; The iris template matching includes comparing the real-time collected and processed iris template with a reference iris template, calculating the feature distance deviation between the two, and determining whether they are the iris templates of the same user, when the feature distance deviation reaches the second preset threshold, it is determined that the matching is successful, indicating that they are the iris templates of the same user, and the identity authentication is completed, when the feature distance deviation does not reach the second preset threshold, it is determined that the matching fails; The reference iris template includes an initial iris template determined by the user in the registration stage of the iris template multi-level life cycle management system.

9. The iris template multi-level life cycle management and dynamic drift eviction algorithm of claim 3, wherein, The module layer, the edge layer and the cloud layer achieve hierarchical security protection through physical isolation; The module layer is used as a cache layer; The edge layer is used as an intermediate cache layer; The cloud layer is used as a permanent storage layer; The overall comparison delay time of the three-layer hierarchical storage architecture is controlled within the eighth unit time.

10. An algorithm for iris template multi-level life cycle management and dynamic drift eviction, characterized by, The steps include: Step S100: real-time acquisition of comparison data of the iris template; Step S200: calculation of drift accumulation through accumulated feature distance deviation, and calculation of user activity through an exponential decay model; Step S300: combination of risk factors to evaluate the state of the iris template; Step S400: execution of hierarchical management strategy according to the drift accumulation; Step S500: execution of dynamic elimination operation according to the hierarchical management strategy.

Images