A secure verification method and system for personal identification codes in the financial sector

By constructing a feature library and combining hash encryption with a recurrent neural network, the security vulnerabilities in traditional personal identification code verification are solved, enabling dynamic identity verification and device adaptation, thus improving the security and applicability in the financial sector.

CN123116253APending Publication Date: 2026-11-17SHENZHEN INSIGMA INNOVATION SOFTWARE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611182141.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-05
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional personal identification number (PIN) verification systems suffer from problems such as weak passwords, plaintext leakage, easy theft and reuse of fixed credentials, lack of anomaly protection, and poor algorithm-hardware compatibility.

Method used

By acquiring user identity information to build a feature library, collecting custom content for security verification, using hash encryption to generate encrypted identification codes, combining recurrent neural networks to generate dynamic prompt data packets for identity verification, and locking the device to reset credentials when consecutive failures occur, a security closed loop is formed.

Benefits of technology

It achieves secure verification of user identity, prevents plaintext leakage and credential theft and reuse, and is compatible with multiple types of terminal devices, thus improving security and adaptability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116253A_ABST
    Figure CN123116253A_ABST
Patent Text Reader

Abstract

This invention discloses a secure verification method and system for personal identification codes in the financial field, relating to the technical field of information security. It involves constructing a feature library, collecting a basic configuration set, verifying the security level of hidden identification codes using the feature library, generating encrypted identification codes through hash encryption, and establishing a standardized prompt matching library; training a recurrent neural network; when a user initiates authentication, the network outputs a single-use dynamic prompt data packet; the user calculates and enters a temporary dynamic password based on the prompt, and the system reverse-engineers the identification code and compares it with the hash ciphertext to complete the verification; if three consecutive verification failures occur, the device is locked, a verification code is issued to reset the configuration, and the new configuration re-enters the security verification process; by screening user information and using irreversible encryption to protect the core identification code, relying on the neural network to generate differentiated authentication prompts, and using a temporary dynamic password to complete identity comparison, the system avoids the risk of leakage and reuse of original credentials, and includes an abnormal locking and reset mechanism to enhance security protection capabilities.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of information security technology, specifically to a secure verification method and system for personal identification codes in the financial sector. Background Technology

[0002] In financial transactions, personal identification numbers (PINs) are the most common identity verification method used by ATMs, POS terminals, mobile banking, and digital wallets for account login, fund transfers, and payments. Leveraging their ease of memorization and simple interaction, short digital codes are widely adopted by various financial institutions globally. Currently, the industry predominantly uses static PIN verification systems, which have revealed several inherent flaws in practical application.

[0003] There is a lack of standardized screening mechanisms when users set identification codes. Most users choose birthdays, mobile phone numbers, bank card fragments, or consecutive or repeating numbers as identification codes. These weak number combinations lack security protection capabilities and can be easily cracked using dictionary attacks or brute-force methods. At the same time, existing storage methods often have vulnerabilities in storing plaintext temporarily. Even if the database is simply encrypted, there is a lack of a mechanism to clear the plaintext cache immediately after processing. Once the stored data is leaked, the user's core credentials will be directly exposed.

[0004] Existing traditional identification code authentication systems are prone to problems such as weak credential settings and leakage of original credential plaintext storage. During authentication interaction, fixed credentials are easily captured by various side-channel methods and can be repeatedly stolen after capture. There is a lack of a differentiated temporary interaction credential generation mechanism and no corresponding anomaly interception and reset closed loop. At the same time, the relevant authentication algorithms have high computing power overhead, rely on dedicated hardware, and are difficult to adapt to various types of financial smart terminal devices. Summary of the Invention

[0005] The purpose of this invention is to address the multiple security and implementation defects of traditional PIN authentication mentioned in the background art, such as weak passwords, plaintext leakage, easy theft and reuse of fixed credentials, lack of anomaly protection, and poor algorithm hardware adaptability. In order to propose a secure verification method and system for personal identification codes in the financial field, this invention provides a solution for such verification.

[0006] A first aspect of this invention provides a secure verification method for personal identification codes in the financial field, the method comprising: A feature library is constructed by obtaining the target user's identity information; The system collects the target user's custom content to obtain a basic configuration set. It then performs a security check on the hidden identification code in the basic configuration set based on the feature library. If the check is successful, it uses hash encryption to encrypt the hidden identification code and obtains an encrypted identification code and a prompt pairing library with the feature library. Obtain the target user's historical session data, and train a recurrent neural network model based on the prompt matching library and the historical session data; When the target user authenticates, the recurrent neural network model generates a single-use dedicated dynamic prompt data packet; The target user inputs a single dynamic password according to the single dedicated dynamic prompt data packet. The system receives the dynamic password, calculates the decryption identification code, and compares the decryption identification code with the encryption identification code to complete the identity verification. If identity verification fails three times in a row, the device will be locked and a reset verification code will be issued for resetting. After resetting, a new configuration set will be generated and the device will re-enter the strength verification process.

[0007] The feature database built on user identity is used to verify the hidden identification code. The credentials are stored with hash encryption to avoid plaintext leakage. Then, a recurrent neural network is trained using historical sessions and a matching database to generate independent dynamic prompts for each time. The temporary dynamic password is used to complete the reverse comparison to prevent the theft and reuse of credentials. Combined with the authentication failure lock and reset mechanism, the security loop is perfected. The overall operation is lightweight and does not require dedicated hardware. It can be adapted to various terminals and solves the multiple security and implementation defects of traditional identification code authentication.

[0008] Optionally, the feature library constructed by obtaining the target user's identity information includes: The identity information of the target user is determined, digital fragments of the identity information are extracted, and an index relationship is established based on the digital fragments to obtain a feature library.

[0009] Optionally, performing security verification on the hidden identifier in the basic configuration set based on the feature library includes: Extract the hidden personal identification code from the basic configuration set, perform basic rule detection on the hidden personal identification code, and if it is qualified, perform identity association detection; otherwise, return the user to reset the hidden personal identification code. The basic rule detection is used to detect whether the hidden personal identification code has duplicate numbers or consecutive sequential numbers. The hidden personal identification code is subjected to identity association detection. If it passes the test, the security score of the hidden personal identification code is calculated. Otherwise, the user is prompted to reset the hidden personal identification code. The identity association detection is used to compare the hidden personal identification code with all mobile phone numbers, birthdays, and bank card number fragments in the feature database to determine whether the identification code is derived from the user's personal information. The security score of the hidden personal identification code is calculated using the Hamming difference metric. If the security score is greater than the security threshold, the code is deemed qualified; otherwise, the user is asked to reset the hidden personal identification code.

[0010] Extracting user identity data fragments and building an index to construct a feature library, then sequentially performing basic detection of repeated or continuous numbers on the hidden identification code and correlation detection by comparing it with the identity data, and finally calculating a security score through Hamming difference metric to complete graded verification, can intercept various weak identification codes that are easy to guess in advance, reducing the probability of brute force and dictionary attacks from the source.

[0011] Optionally, the encrypted identification code and the hint pairing library obtained by hashing the hidden identification code and matching it with the feature library include: Read the hidden identification code temporarily loaded in memory, and call the system's built-in NIST standard KMAC key hash algorithm module to load the encryption key; The hidden identification code corresponding to the numeric string is used as the original input data. The numeric string and the encryption key are input into the KMAC operation interface to perform multiple rounds of KECCAK permutation and iteration operations to generate the authentication ciphertext. Immediately after the operation is completed, the plaintext cache of the hidden identification code in memory is cleared, and the authentication ciphertext is separately encapsulated and marked to obtain the encrypted identification code; Traverse all original prompt operator pairing groups in the feature library, extract the user-defined prompt and addition / subtraction number operator in each group, check for duplicates, and bind the user-defined prompt with the corresponding addition / subtraction number operator to obtain standard pairing entries; All standard pairing entries are sorted in ascending order according to the preset encoding order of the custom prompts, and a unique numerical index number is assigned to each pairing entry to build an index and obtain the prompt pairing library.

[0012] The hidden identification code is encrypted by using the standard KMAC algorithm combined with multiple rounds of KECCAK permutation operations. After the operation, the plaintext cache is cleared in time to generate the encrypted identification code. At the same time, the original prompt operator pairing groups are traversed to remove duplicates, standardize the binding, and build an index to form a prompt pairing library. This not only avoids the risk of leakage of the original identification code plaintext, but also generates a well-structured standardized dataset that can be directly called by the model.

[0013] Optionally, training a recurrent neural network model based on the prompt matching library and the historical session data includes: Read the prompt data index from the prompt pairing library, read the session data index from the historical session data, concatenate the prompt data index and the session data index to obtain fused data, and sort the fused data in chronological order to obtain the training dataset; The basic structure of a recurrent neural network is built, and the parameters of the recurrent neural network are initialized to obtain a recurrent neural network model. The training dataset is then input into the recurrent neural network model for training to obtain a trained recurrent neural network model.

[0014] Optionally, the system receives the decryption identification code obtained from the dynamic password calculation, including: Read the dynamic password stored in this session, extract the numeric operators and random input empty space path from the dynamic password, read the single valid dynamic password entered by the user, and extract the numeric content filled in the random empty space; According to the random input empty space path, determine the operator operation relationship corresponding to each number in the random input empty space path, perform the operator inverse operation on each filled number in the second valid dynamic password, fill in the empty and unfilled number positions in the random path, and obtain the decryption identification code.

[0015] A second aspect of this invention provides a secure verification system for personal identification codes in the financial sector, the system comprising: The feature library construction module is used to obtain the target user's identity information and construct a feature library. The security verification module is used to collect the target user's custom content to obtain a basic configuration set, and to perform security verification on the hidden identification code in the basic configuration set according to the feature library. If it is qualified, the hidden identification code is encrypted by hashing and matched with the feature library to obtain an encrypted identification code and a prompt pairing library. The model training module is used to acquire the target user's historical session data and train a recurrent neural network model based on the prompt matching library and the historical session data. A data packet generation module is used to generate a single-use dynamic prompt data packet by the recurrent neural network model when the target user performs authentication. The password comparison module is used for the target user to input a single dynamic password according to the single dedicated dynamic prompt data packet. The system receives the dynamic password, calculates the decryption identification code, and compares the decryption identification code with the encryption identification code to complete the identity verification. The verification and locking module is used to lock the device and issue a reset verification code to reset it if the identity verification fails three times in a row. After the reset, a new configuration set is generated and the device re-enters the strength verification process.

[0016] The beneficial effects of this invention are: This invention proposes a secure verification method for personal identification codes in the financial sector. It screens weak identification codes using user identity data and then encrypts and stores core credentials using an irreversible hashing method, eliminating the risk of plaintext leakage. By leveraging a recurrent neural network to generate unique authentication prompts each time, users only need to enter a temporary dynamic password to complete verification. Even if the interaction process is spied on or recorded, the stolen temporary credentials cannot be reused, resisting various side-channel attacks. Simultaneously, it establishes a secure closed loop by intercepting authentication anomalies and implementing an account reset process. The overall computation is lightweight, requiring no additional dedicated hardware and adaptable to various financial and smart terminal devices. While achieving comprehensive identity protection, it also considers device compatibility and practical feasibility for implementation. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a secure verification method for personal identification codes in the financial field, provided as an embodiment of the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0019] This invention provides a secure verification method for personal identification codes in the financial sector. See also... Figure 1 The method includes the following steps: S101, Obtain the target user's identity information to construct a feature library; S102: Collect the target user's custom content to obtain a basic configuration set. Perform security verification on the hidden identification code in the basic configuration set according to the feature library. If it is qualified, obtain the encrypted identification code and prompt pairing library by hashing the hidden identification code and the feature library. S103, Obtain the target user's historical session data, and train a recurrent neural network model based on the prompt matching library and historical session data; S104, When the target user authenticates, the recurrent neural network model generates a single-use dedicated dynamic prompt data packet; S105, the target user enters a one-time dynamic password according to the one-time dedicated dynamic prompt data packet, the system receives the dynamic password, calculates the decryption identification code, and compares the decryption identification code with the encryption identification code to complete the identity verification; S106 If identity verification fails three times in a row, the device will be locked and a reset verification code will be issued for resetting. After resetting, a new configuration set will be generated and the strength verification process will be restarted.

[0020] In one implementation, a basic configuration set is obtained by collecting the target user's custom content: the custom content includes the user-set hidden identification code, simple addition and subtraction prompt operators, and account reset channels; In one implementation, the target user enters a one-time dynamic password based on a dedicated dynamic prompt data packet. The aforementioned operator only involves simple addition and subtraction of numbers such as ±1, ±2, and ±5, and is not a complex mathematical operation. When registering, the user should memorize the simple addition and subtraction rules corresponding to each set of prompts in advance. During authentication, the user should refer to the prompts and perform simple addition and subtraction on each digit of their fixed hidden PIN to complete the process.

[0021] In one implementation, the system receives a decryption identification code obtained by dynamic password calculation, and compares the decryption identification code with the encryption identification code to complete identity verification: calling the KMAC encryption module that is completely consistent with the registration stage and loading the same set of financial special encryption keys; The derived hidden identification code string is used as input data and fed into the KMAC operation interface along with the encryption key; KECCAK permutation iteration operation is performed to generate the real-time hash ciphertext corresponding to the deduced identification code; after the operation is completed, the hidden identification code plaintext temporarily cached in memory is cleared immediately; The backend only stores irreversible hash ciphertext and has no original identification code for direct matching. It must be re-hashed to complete the comparison. The encryption standard is completely consistent during the generation and registration stages to ensure that the comparison results are accurate and effective. Read the compliant encrypted hidden personal identification code ciphertext that is persistently stored in the database for this user; The hash ciphertext is derived in real time by comparing each bit with the hash ciphertext stored in the background. If the two encrypted messages are completely identical, the identity verification is considered successful; if the encrypted messages differ, the authentication is considered to have failed, and the cumulative error count is incremented by 1; the identity verification result is (pass / fail). It cannot directly verify the legitimacy of dynamic passwords, nor can it confirm that the operator possesses a legitimate hidden identification code. However, it can accurately verify the legitimacy of user identity without disclosing or persistently storing the original hidden identification code.

[0022] In one implementation, the system continuously accumulates the number of authentication failures. When three consecutive hash-ciphertext comparisons fail, the system immediately locks the authentication interaction entry point of the current device and simultaneously retrieves the reserved user's email address or mobile phone number to send a one-time reset verification code. After the user completes the verification code verification, they can reset a new hidden personal identification code, hint symbols, and numerical operators to generate a new basic hidden personal identification code configuration set. This new configuration set will then re-perform the entire process of hidden personal identification code security strength testing, encryption, and hint matching library construction.

[0023] In one embodiment, obtaining the target user's identity information to construct a feature library includes: Identify the target user's identity information, extract digital fragments of the identity information, and establish an index relationship based on the digital fragments to obtain a feature library.

[0024] In one implementation, the user fills in original identity information such as name, mobile phone number, ID card number, birthday, bank card number, and family postal code; the system inputs and stores all original identity information text; the output product is a set of original identity information. Since the user's basic personal information is lacking, it is impossible to extract digital material used to verify the hidden personal identification code. Therefore, obtaining the user's identity information provides the original data source for extracting digital fragments.

[0025] In one implementation, the original set of identity information is processed by extracting consecutive 4-digit and 6-digit number segments from each identity information, removing duplicate segments, and uniformly organizing them into number units. Since the original set of user identity number segments contains a mixture of text and numbers, it is impossible to directly compare and screen weak passwords with the hidden personal identification code. Therefore, all number strings related to the user are separated and purified, and only the number content that can be used for password comparison is retained.

[0026] In one implementation, the original set of user identity digital fragments is uniformly classified, deduplicated, and sorted through data-driven modeling to establish an index mapping relationship; a standardized user identity digital feature library is constructed, as the retrieval efficiency of scattered and disordered digital fragments is low, which will lead to excessively long verification time for the hidden personal identification code; a structured and fast-searchable feature library is constructed, and subsequent steps can quickly compare whether the hidden personal identification code contains user identity digital data.

[0027] In one embodiment, performing security verification on the hidden identifier in the basic configuration set based on the feature library includes: Extract the hidden personal identification code from the basic configuration set, perform basic rule checks on the hidden personal identification code, and if it passes, perform identity association checks; otherwise, return the user to reset the hidden personal identification code. The basic rule checks are used to detect whether the hidden personal identification code contains duplicate numbers or consecutive numbers. The hidden personal identification code is subjected to identity association detection. If it passes the test, the security score of the hidden personal identification code is calculated. Otherwise, the user is returned to reset the hidden personal identification code. The identity association detection is used to compare the hidden personal identification code with all mobile phone numbers, birthdays, and bank card number fragments in the feature database to determine whether the identification code is derived from the user's personal information. The security score of the Hidden Personal Identifier (HPI) is calculated using the Hamming Difference Measure. If the security score is greater than the security threshold, the user is deemed qualified; otherwise, the user is asked to reset their HPI.

[0028] In one implementation, a basic hidden personal identification code configuration set and a standardized user identity digital feature library are used to extract hidden personal identification codes from the basic hidden personal identification code configuration set. The configuration set is a mixed package of data, which cannot be used for security verification, encryption, and pairing and sorting on its own. It is necessary to separate the core data into two independent and processable types and perform security detection and pairing and sorting separately.

[0029] In one implementation, the basic rule verification involves checking whether the hidden personal identification code contains duplicate or consecutive numbers; the identity association verification involves comparing the hidden personal identification code with all mobile phone numbers, birthdays, and bank card number fragments in the feature database to determine whether the identification code is derived from the user's personal information; and the password security score is calculated using the Hamming difference metric. If the security score does not reach the threshold, a modification guide is pushed to the user, returning the user to reset the hidden personal identification code; otherwise, the hidden personal identification code is deemed secure and qualified. Users can easily set weak personal identification codes such as birthdays or consecutive numbers, which are easily cracked by dictionary attacks and brute-force guessing; filtering all high-risk weak identification codes reduces the probability of the identification code being guessed.

[0030] In one embodiment, obtaining the encrypted identification code and hint pairing library through hash-encrypted hidden identification code and feature library includes: Read the hidden identification code temporarily loaded in memory, and call the system's built-in NIST standard KMAC key hash algorithm module to load the encryption key; The numeric string corresponding to the hidden identification code is used as the original input data. The numeric string and the encryption key are input into the KMAC operation interface to perform multiple rounds of KECCAK permutation and iteration operations to generate the authentication ciphertext. After the operation is completed, the plaintext cache of the hidden identification code in memory is immediately cleared, and the authentication ciphertext is separately encapsulated and marked to obtain the encrypted identification code; Traverse all original prompt operator pairing groups in the feature library, extract the user-defined prompt and addition / subtraction number operator in each group, check for duplicates, and bind the user-defined prompt with the corresponding addition / subtraction number operator to obtain standard pairing entries. All standard pairing entries are sorted in ascending order according to the preset encoding order of the custom prompts, and a unique numerical index number is assigned to each pairing entry to build an index and obtain the prompt pairing library.

[0031] In one implementation, the system reads the secure, qualified hidden personal identification code plaintext temporarily loaded in memory without writing it to local persistent storage; it calls the system's built-in NIST standard KMAC key hashing algorithm module to load a financial scenario-specific encryption key; it uses the hidden personal identification code numeric string as the original input data, and passes it along with the encryption key to the KMAC operation interface; it performs multiple rounds of KECCAK permutation iteration operations to generate a fixed-length irreversible message authentication ciphertext; immediately after the operation is completed, it clears the hidden personal identification code plaintext cache in memory; and it separately encapsulates and marks the generated ciphertext to distinguish different user identities. Storing hidden personal identification codes in plaintext poses a direct risk of user funds theft if the database is leaked. Storing only irreversible encrypted data ensures that even if the data is leaked, attackers cannot reverse-engineer the original hidden personal identification codes.

[0032] In one implementation, all original prompt operator pairs are traversed, and user-defined prompt symbols and system-assigned addition and subtraction operators are extracted from each pair. Filter and remove invalid pairs that contain duplicate prompt symbols, multiple operators bound to the same prompt, or duplicate bindings of the same operator; Each of the remaining unique prompt symbols is fixedly bound to its corresponding exclusive numerical operator to form a standard pairing entry; Unified sorting: All standard pairing entries are sorted in ascending order according to the preset coding order of the prompt symbols; Establish index mapping: Assign a unique numerical index number to each paired entry, and build a two-way lookup mapping table between the index and hints / operators; All indexed standard pairing entries are uniformly encapsulated to generate a structured dataset file (hint pairing library); the original hint operator pairing groups are messy and unindexed, making them unreadable for subsequent neural network training; the output is a structured pairing dataset that can be directly input into the model for training.

[0033] In one embodiment, training a recurrent neural network model based on a prompting matching library and historical session data includes: Read the prompt data index from the prompt pairing library, read the session data index from the historical session data, concatenate the prompt data index and the session data index to obtain the fused data, and sort the fused data in chronological order to obtain the training dataset; The basic structure of a recurrent neural network is built, and the parameters of the recurrent neural network are initialized to obtain a recurrent neural network model. The training dataset is then input into the recurrent neural network model for training to obtain a trained recurrent neural network model.

[0034] In one implementation, all standardized prompt symbols, corresponding numeric operators, and unique index numbers in the prompt operator pairing library are read; historical authentication session state data is read, and prompt entries, random input paths, and time sequence markers used in each authentication are extracted. The two types of data are fused and spliced ​​together, and sorted according to the authentication time sequence to construct a complete training dataset with time sequence labels (time sequence fusion training dataset, also known as training dataset). The pairing library and historical session data are independent of each other and have no time sequence correlation, so they cannot be used as input materials for the time sequence neural network. Therefore, complete training samples with chronological order are generated first to meet the input requirements of the recurrent neural network for time sequence learning.

[0035] In one implementation, a lightweight wide-learning recurrent neural network infrastructure is built, and all original parameters such as network weights, biases, wide-learning random mapping layers, and temporal state storage units are initialized; training constraints are set: prompts and random input paths in two adjacent outputs cannot be repeated; the temporal fusion training dataset is divided into a training set and a validation set, which are then input into the initial network.

[0036] In one implementation, the training dataset is input into the recurrent neural network in batches according to the time series. Based on the input prompts and historical session time series data, the network predicts the prompt items and input paths that can be output in the next round. The prediction results are compared with the constraint rules (the prompts cannot be repeated), and the prediction bias loss value is calculated. Based on the loss value, backpropagation is performed to correct all parameters to be updated, such as network weights, biases, wide learning mapping coefficients, and temporal memory unit weights, layer by layer. The iterative execution process is as follows: Input sample → Prediction → Calculate loss → Update parameters. Each iteration completes a set of parameter updates. After each iteration, the model output is verified using a validation set. If the loss no longer decreases after multiple consecutive validation iterations, the parameters are considered to have converged, and the iteration is stopped. This completes the optimization of a new set of network parameters after multiple iterations. The initial model parameters are random, and the output prompts and input paths have repeated vulnerabilities, making it easy for attackers to repeatedly record the screen and piece together the personal identification code. We continuously optimize the network parameters to eliminate the defects of continuous authentication prompts and repeated paths. The converged set of optimized parameters is fully loaded into the initial network framework, the network structure and parameters are fixed, and the iterative training channel is closed.

[0037] In one embodiment, the system receives a decryption identification code obtained through dynamic cryptographic calculation, including: Read the dynamic password stored in this session, extract the numeric operators and random input blank paths from the dynamic password, read the single-valid dynamic password entered by the user, and extract the numeric content filled in the random blanks. Based on the random input empty space path, determine the operator operation relationship corresponding to each number in the random input empty space path, perform the operator inverse operation on each filled number in the second valid dynamic password, fill in the empty and unfilled number positions in the random path, and obtain the decryption identification code.

[0038] In one implementation, the system retrieves the single-authentication dynamic prompt data packet stored in the current session and extracts the effective numerical operator and random input space path rule from it; it reads the single-valid dynamic password entered by the user and extracts the numerical content filled in the random space; since the operator and path rule corresponding to this conversion are missing, it is impossible to reverse deduce the original hidden personal identification code. All the temporary session parameters required for reverse restoration are collected to provide a calculation basis for restoring the hidden identification code.

[0039] In one implementation, the operator operation relationship corresponding to each digit of the hidden personal identification code is located according to the randomly input empty space path; Perform inverse operator operations on each digit entered in the dynamic password (if it was added during registration, subtract it this time; if it was subtracted during registration, add it this time). Fill in the missing numbers in the random path and concatenate them to generate a complete and qualified secure hidden personal identification code plaintext. After the operation is completed, the plaintext identification code is only temporarily cached and not written to the database for persistent storage. The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention should still fall within the scope of the claims of the present invention.

Claims

1. A secure verification method for personal identification codes in the financial sector, characterized in that, The method includes: A feature library is constructed by obtaining the target user's identity information; The system collects the target user's custom content to obtain a basic configuration set. It then performs a security check on the hidden identification code in the basic configuration set based on the feature library. If the check is successful, it uses hash encryption to encrypt the hidden identification code and obtains an encrypted identification code and a prompt pairing library with the feature library. Obtain the target user's historical session data, and train a recurrent neural network model based on the prompt matching library and the historical session data; When the target user authenticates, the recurrent neural network model generates a single-use dedicated dynamic prompt data packet; The target user inputs a single dynamic password according to the single dedicated dynamic prompt data packet. The system receives the dynamic password, calculates the decryption identification code, and compares the decryption identification code with the encryption identification code to complete the identity verification. If identity verification fails three times in a row, the device will be locked and a reset verification code will be issued for resetting. After resetting, a new configuration set will be generated and the device will re-enter the strength verification process.

2. The secure verification method for personal identification codes in the financial field according to claim 1, characterized in that, The feature library constructed by obtaining the target user's identity information includes: The identity information of the target user is determined, digital fragments of the identity information are extracted, and an index relationship is established based on the digital fragments to obtain a feature library.

3. The secure verification method for personal identification codes in the financial field according to claim 1, characterized in that, The security verification of the hidden identifier in the basic configuration set based on the feature library includes: Extract the hidden personal identification code from the basic configuration set, perform basic rule detection on the hidden personal identification code, and if it is qualified, perform identity association detection; otherwise, return the user to reset the hidden personal identification code. The basic rule detection is used to detect whether the hidden personal identification code has duplicate numbers or consecutive sequential numbers. The hidden personal identification code is subjected to identity association detection. If it passes the test, the security score of the hidden personal identification code is calculated. Otherwise, the user is prompted to reset the hidden personal identification code. The identity association detection is used to compare the hidden personal identification code with all mobile phone numbers, birthdays, and bank card number fragments in the feature database to determine whether the identification code is derived from the user's personal information. The security score of the hidden personal identification code is calculated using the Hamming difference metric. If the security score is greater than the security threshold, the code is deemed qualified; otherwise, the user is asked to reset the hidden personal identification code.

4. The secure verification method for personal identification codes in the financial field according to claim 1, characterized in that, The encrypted identification code and the hint pairing library obtained by hashing the hidden identification code and matching it with the feature library include: Read the hidden identification code temporarily loaded in memory, and call the system's built-in NIST standard KMAC key hash algorithm module to load the encryption key; The hidden identification code corresponding to the numeric string is used as the original input data. The numeric string and the encryption key are input into the KMAC operation interface to perform multiple rounds of KECCAK permutation and iteration operations to generate the authentication ciphertext. Immediately after the operation is completed, the plaintext cache of the hidden identification code in memory is cleared, and the authentication ciphertext is separately encapsulated and marked to obtain the encrypted identification code; Traverse all original prompt operator pairing groups in the feature library, extract the user-defined prompt and addition / subtraction number operator in each group, check for duplicates, and bind the user-defined prompt with the corresponding addition / subtraction number operator to obtain standard pairing entries; All standard pairing entries are sorted in ascending order according to the preset encoding order of the custom prompts, and a unique numerical index number is assigned to each pairing entry to build an index and obtain the prompt pairing library.

5. A secure verification method for personal identification codes in the financial field according to claim 1, characterized in that, Training a recurrent neural network model based on the prompt matching library and the historical session data includes: Read the prompt data index from the prompt pairing library, read the session data index from the historical session data, concatenate the prompt data index and the session data index to obtain fused data, and sort the fused data in chronological order to obtain the training dataset; The basic structure of a recurrent neural network is built, and the parameters of the recurrent neural network are initialized to obtain a recurrent neural network model. The training dataset is then input into the recurrent neural network model for training to obtain a trained recurrent neural network model.

6. A secure verification method for personal identification codes in the financial field according to claim 1, characterized in that, The system receives the decryption identification code obtained from the dynamic password calculation, which includes: Read the dynamic password stored in this session, extract the numeric operators and random input empty space path from the dynamic password, read the single valid dynamic password entered by the user, and extract the numeric content filled in the random empty space; According to the random input empty space path, determine the operator operation relationship corresponding to each number in the random input empty space path, perform the operator inverse operation on each filled number in the second valid dynamic password, fill in the empty and unfilled number positions in the random path, and obtain the decryption identification code.

7. A secure verification system for personal identification codes in the financial sector, characterized in that, The system includes: The feature library construction module is used to obtain the target user's identity information and construct a feature library. The security verification module is used to collect the target user's custom content to obtain a basic configuration set, and to perform security verification on the hidden identification code in the basic configuration set according to the feature library. If it is qualified, the hidden identification code is encrypted by hashing and matched with the feature library to obtain an encrypted identification code and a prompt pairing library. The model training module is used to acquire the target user's historical session data and train a recurrent neural network model based on the prompt matching library and the historical session data. A data packet generation module is used to generate a single-use dynamic prompt data packet by the recurrent neural network model when the target user performs authentication. The password comparison module is used for the target user to input a single dynamic password according to the single dedicated dynamic prompt data packet. The system receives the dynamic password, calculates the decryption identification code, and compares the decryption identification code with the encryption identification code to complete the identity verification. The verification and locking module is used to lock the device and issue a reset verification code to reset it if the identity verification fails three times in a row. After the reset, a new configuration set is generated and the device re-enters the strength verification process.