A multi-strategy fusion password detection method, system and medium

Abstract

The application provides a multi-strategy fusion password detection method and system and a medium, and relates to the technical field of data processing.The method comprises the following steps: after receiving an initial registration password submitted by a user, calculating a lightweight hash; performing weak password label hit matching and outputting a password state matching result; if no hit is found, extracting a parallel feature vector through an SIMD instruction set; loading the parallel feature vector into a multi-strategy fusion model to perform multi-strategy password strength characteristic fusion evaluation and output a real-time password strength label; matching an asynchronous scene to perform asynchronous double-thread application scene risk detection and output a heterogeneous scene risk report; randomly disturbing the initial registration password to obtain a temporary backup password library; and loading the temporary backup password library into a registration interface for visual display.The application solves the technical problem that the prior art relies on a model with a large amount of calculation to perform password evaluation, which results in slow evaluation speed, cannot realize real-time password detection, and further affects user experience and system efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a multi-strategy fusion password detection method and system and a medium. BACKGROUND

[0002] As a common identity verification method, passwords directly affect the security of user accounts and data. However, many existing password detection methods have some shortcomings and cannot effectively deal with complex attack methods. On the one hand, existing technologies rely on a single password evaluation strategy, which usually evaluates passwords by detecting their length, character types, or compliance with preset rules. Although this method can effectively detect some common passwords, it cannot identify complex passwords or new password variants, resulting in some complex but non-compliant passwords not being detected, increasing the risk of password cracking. On the other hand, existing technologies rely on computationally intensive models for password evaluation, which results in slow evaluation speed, especially in large-scale user registration and login scenarios, making real-time or near-real-time evaluation impossible, thereby affecting user experience and system efficiency. SUMMARY

[0003] The present application provides a multi-strategy fusion password detection method, system and medium, aiming to solve the technical problem that existing technologies rely on computationally intensive models for password evaluation, resulting in slow evaluation speed and inability to achieve real-time password detection, thereby affecting user experience and system efficiency.

[0004] The first aspect of the present application provides a multi-strategy fusion password detection method, which comprises: after receiving an initial registration password submitted by a user, calculating a lightweight hash of the initial registration password; using the lightweight hash to perform weak password rule library hit matching, and outputting a password state matching result, wherein the weak password rule library is preloaded when the user IU interface is switched to the registration interface; if the password state matching result is a miss, extracting a parallel feature vector of the initial registration password through a SIMD instruction set; loading the parallel feature vector into a pre-constructed multi-strategy fusion model, performing multi-strategy password strength characteristic fusion evaluation via the multi-strategy fusion model, and outputting a real-time password strength label; matching asynchronous scenarios according to the real-time password strength label, performing asynchronous double-thread application scenario risk detection on the initial registration password, and outputting a heterogeneous scenario risk report; randomly perturbing the initial registration password according to the heterogeneous scenario risk report to obtain a temporary backup password library; and loading the temporary backup password library into the registration interface for visual display.

[0005] In a second aspect, the application discloses a multi-strategy fusion password detection system, which is used for the multi-strategy fusion password detection method, and comprises the following modules: a lightweight hash calculation module, which is used for calculating a lightweight hash of an initial registration password after receiving the initial registration password submitted by a user; a label hit matching module, which is used for performing weak password label hit matching on a weak password rule library by using the lightweight hash, and outputting a password state matching result, wherein the weak password rule library is preloaded when a user IU interface is switched to a registration interface; a parallel feature vector extraction module, which is used for extracting a parallel feature vector of the initial registration password by using a SIMD instruction set if the password state matching result is a miss; a strength characteristic fusion evaluation module, which is used for loading the parallel feature vector into a pre-constructed multi-strategy fusion model, performing multi-strategy password strength characteristic fusion evaluation by using the multi-strategy fusion model, and outputting a real-time password strength label; an application scenario risk detection module, which is used for performing asynchronous double-thread application scenario risk detection on the initial registration password according to the real-time password strength label and an asynchronous scenario, and outputting a heterogeneous scenario risk report; a random disturbance module, which is used for randomly disturbing the initial registration password according to the heterogeneous scenario risk report, and obtaining a temporary backup password library; and a backup password library loading module, which is used for loading the temporary backup password library to a registration interface for visual display.

[0006] In a third aspect, the application discloses a storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of the multi-strategy fusion password detection method.

[0007] The one or more technical solutions provided in the application have at least the following beneficial effects:

[0008] By calculating the lightweight hash of the initial registration password, the processing efficiency of password verification can be greatly improved. The lightweight hash algorithm is faster and occupies less resources than the traditional hash algorithm, reducing the load of the system while ensuring the security of password processing. This optimization enables password detection to respond quickly and handle large-scale user-submitted passwords, improving the real-time performance and response speed of the system. By using a preloaded weak password rule library to quickly match the lightweight hash, weak passwords can be efficiently identified and intercepted. This matching process not only effectively prevents users from using easily cracked passwords, but also ensures the efficiency and accuracy of password detection, further improving detection efficiency. By using the SIMD (Single Instruction Multiple Data) instruction set for parallel processing of password feature extraction, the complex feature extraction process of passwords can be accelerated. The SIMD instruction set allows a computer to process multiple data elements simultaneously in one operation, significantly improving processing speed. This technology improves the performance and efficiency of password detection, ensuring that the system can quickly process user-submitted passwords. The multi-strategy fusion model combines different evaluation strategies to more accurately and comprehensively evaluate password strength, enhancing the intelligence and diversity of password detection. The multi-strategy fusion model outputs real-time password strength labels, enabling real-time response to user input and improving user experience. By using asynchronous dual-thread application scenario risk detection, multiple password risk detection tasks can be executed simultaneously, which not only improves detection efficiency but also avoids performance bottlenecks caused by multi-task execution. The application of the dual-thread mode enables the system to complete comprehensive evaluation of passwords in the shortest time, providing timely security feedback to users. By randomly perturbing the initial registration password to generate a temporary backup password library, the diversity of passwords can be further enhanced, avoiding the use of simple or known password combinations. This provides an intelligent way for the system to deal with weak password selection by users, effectively increasing password complexity and unpredictability and improving account security. Loading the temporary backup password library into the registration interface for visual display provides an interactive way for users to view and select passwords. This display method helps users easily select stronger passwords, further enhancing password security management.

[0009] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the following specific embodiments of the present application can be implemented according to the content of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 A multi-strategy fusion password detection method flowchart provided by an embodiment of the present application.

[0011] Figure 2A multi-strategy fusion password detection system structure schematic diagram provided by an embodiment of the application.

[0012] Marked: lightweight hash calculation module 10, mark hit matching module 20, parallel feature vector extraction module 30, strength characteristic fusion evaluation module 40, application scenario risk detection module 50, random disturbance module 60, backup password library loading module 70. DETAILED DESCRIPTION

[0013] The embodiment of the application provides a multi-strategy fusion password detection method, system and medium, and solves the technical problem that the prior art depends on a model with a large amount of calculation to perform password evaluation, which leads to slow evaluation speed, cannot realize real-time password detection, and further affects user experience and system efficiency.

[0014] After introducing the basic principle of the application, various non-limiting embodiments of the application will be specifically introduced in combination with the drawings of the specification. It should be understood that the specific embodiments described herein are only used to explain the application and not to limit the application.

[0015] Embodiment one, as shown in the figure, the embodiment of the application provides a multi-strategy fusion password detection method, which comprises: Figure 1

[0016] After receiving the initial registration password submitted by the user, the lightweight hash of the initial registration password is calculated.

[0017] After the user inputs the initial registration password and submits, the initial registration password is received, which is a password created by the user in the registration process, which can be a common weak password or a relatively complex password. In order to ensure the security of the password, the initial registration password is first processed by a lightweight hash algorithm. Hash algorithm is a one-way function that converts the input password into a fixed-length character or number sequence. The hash result is not easy to deduce the original password. Compared with the standard hash algorithm, the lightweight hash algorithm is usually faster and has smaller calculation amount, which avoids complex operation on each password while still effectively ensuring the privacy of the password. The calculated hash value is a unique representation of the password, which can be further used for password detection and comparison without saving the original password itself.

[0018] The lightweight hash is used for weak password rule library to perform weak password mark hit matching, and the password state matching result is output, wherein the weak password rule library is preloaded when the user IU interface is switched to the registration interface.

[0019] ​Preloading when the user IU interface switches to the registration interface, preloading a weak password rule library containing a series of predefined rules for identifying common weak passwords, including a list of common weak passwords, password strength evaluation criteria obtained through cryptographic analysis, and weak password strategies based on common password patterns such as sequential numbers and letters, such as "123456" and "password".

[0020] Matching in the weak password rule library using a lightweight hash to find out if the lightweight hash matches a known weak password hash. If it matches, it means that the initial registration password is a weak password, and the initial registration password is marked as a weak password. The password status matching result is hit, in which case the invalid password is intercepted to avoid invalid passwords entering the subsequent calculation process and reduce computational overhead. If there is no match, it means that the initial registration password does not belong to the common weak password, and the password status matching result is not hit, and the strength of the password is further analyzed.

[0021] If the password status matching result is not hit, the parallel feature vector of the initial registration password is extracted through the SIMD instruction set.

[0022] SIMD (Single Instruction, Multiple Data) is a parallel computing technology that can perform the same operation on multiple data in the same clock cycle. By using the SIMD instruction set, a large amount of data can be efficiently processed and multiple password features can be calculated simultaneously, reducing the calculation time. If the password status matching result is not hit, it means that the initial registration password does not belong to the common weak password. In this case, the features of the initial registration password are extracted into a vector of multiple dimensions using the SIMD instruction set. The vector contains different features of the password, such as character type, character distribution, length, etc. Through SIMD, multiple features of the password can be calculated in parallel in multiple processing units to generate a parallel feature vector. By extracting the parallel feature vector, various information of the initial registration password is encoded into a set of numerical values, which can be further used as input features for password strength evaluation.

[0023] Load the parallel feature vector into a pre-built multi-strategy fusion model and perform multi-strategy password strength feature fusion evaluation via the multi-strategy fusion model to output a real-time password strength label.

[0024] The multi-strategy fusion model is a complex model integrating multiple password strength assessment channels, including a parallel rule engine channel, a machine learning channel, and a strategy adaptation channel, which jointly act to produce more accurate password strength scores by fusing the assessment results of all channels. Through analysis of the parallel feature vectors by the multi-strategy fusion model, real-time password strength labels are output, which reflect the strength level of the password, such as weak password, strong password, etc.

[0025] According to the real-time password strength label matching the asynchronous scenario, the initial registration password is subjected to asynchronous double-thread application scenario risk detection, and a heterogeneous scenario risk report is output.

[0026] According to the real-time password strength label, the initial registration password is placed in different scenarios for evaluation. Specifically, for weak passwords, they are placed in a high-risk double-thread application scenario for parallel processing, including historical password association detection and password leakage state detection, to achieve more stringent security detection. For strong passwords, they can be placed in a low-risk single-thread application scenario, including password leakage state detection, for less detection. Through asynchronous double-thread processing, a heterogeneous scenario risk report is output, reducing the time required for password detection and improving the speed of password detection.

[0027] According to the heterogeneous scenario risk report, the initial registration password is randomly disturbed to obtain a temporary backup password library.

[0028] According to the heterogeneous scenario risk report, if the initial registration password is determined to be unsafe in some scenarios, measures are taken to enhance the security of the password. Specifically, the initial registration password is randomly disturbed, including changing the order of characters, adding random characters, replacing some characters with other characters, etc. In this way, a temporary backup password library is generated, which contains multiple temporarily disturbed backup passwords. These temporary backup passwords have higher complexity and randomness, and therefore are more secure.

[0029] The temporary backup password library is loaded into the registration interface for visual display.

[0030] The generated temporary backup password library is loaded into the user's registration interface and presented to the user through visual display, providing a reference for the user. Through visual display, the user can intuitively select a password that meets the strength requirements, improving the user experience and enhancing the efficiency of password management.

[0031] Further, after receiving the initial registration password submitted by the user, a lightweight hash of the initial registration password is calculated, and the method comprises:

[0032] The user registration information is extracted at the registration interface, and a multi-element associated string is constructed based on preset rules. The initial registration password is expanded using the multi-element associated string to obtain an expanded registration password. The expanded registration password is subjected to multi-scale segmentation to obtain a multi-scale sub-password set. After calculating the multi-scale lightweight hash set of the multi-scale sub-password set, the multi-scale lightweight hash set is accumulated across scales to output a synthetic hash. After calculating the initial hash of the initial registration password, the synthetic hash and the initial hash are weighted based on a preset weight distribution to output the lightweight hash.

[0033] The user registration information, including the username, email prefix, last 4 digits of the mobile phone number, etc., is extracted at the registration interface, and the user registration information is processed through preset rules to form a multi-element associated string. The preset rules include combination or other conversion rules for the user registration information. In this way, the user registration information is converted into an expanded information, which can be used to increase the complexity and uniqueness of the password, making it more difficult to be cracked.

[0034] The expansion processing includes concatenating the multi-element associated string with the initial registration password. For example, if the user's initial password is password123 and the multi-element associated string is example123456, the expanded registration password can be password123example123456. Through expansion processing, the initial registration password can be expanded by adding user registration information, thereby enhancing the complexity of the password and improving its security.

[0035] The expanded registration password is subjected to multi-scale segmentation. Multi-scale means that the segmentation method is performed at different scales. For example, the segmentation can be based on the number of characters, such as every 3 characters as a unit. Or the segmentation can be based on a specific pattern, such as every time a character or number is converted. This multi-scale segmentation splits the expanded registration password into multiple units, each of which is a small part of the original password. Each unit corresponds to different password characteristics and has independent password strength and rules.

[0036] The lightweight hash algorithm is applied to each codon set in the multi-scale sub-codon set for calculation, such as using a relatively simple and efficient hash method, and each codon set obtains a lightweight hash value, thereby obtaining multiple lightweight hash values of different scales, which are associated with different parts and features of the original password and can effectively capture various information of the password. Cross-scale accumulation means combining or accumulating lightweight hash values of different scales, that is, combining the multi-scale lightweight hash set generated by different scale partitions together through direct addition or other accumulation methods to obtain a new composite hash value, i.e., a synthetic hash. This cross-scale accumulation can integrate the multi-level information of each codon set into a hash value, further increasing the complexity and security of the password.

[0037] The initial registration password is hashed to calculate the initial hash of the initial registration password. Unlike multi-scale hashing, the initial hash is a direct hash value calculation of the initial registration password, which is used to represent the basic features of the password itself. The initial hash is a hash value calculated by a standard hash algorithm (such as SHA-256, etc.).

[0038] The synthetic hash and the initial hash are combined using a preset weight. The preset weight is assigned based on password strength, complexity, or other factors. For example, if the synthetic hash contains multiple source information, the synthetic hash is given a higher weight because it represents the characteristics of the password at multiple scales. The initial hash is given a lower weight because it is based on the original password input by the user. By weighted fusion, the synthetic hash and the initial hash are combined by weight to obtain a final hash value, i.e., a lightweight hash. The lightweight hash combines the multi-scale features of the user password and its original password features, making the final hash value have higher complexity and recognition.

[0039] Further, after calculating the multi-scale lightweight hash set of the multi-scale sub-codon set, the multi-scale lightweight hash set is cross-scale accumulated to output a synthetic hash. The method comprises:

[0040] The multi-scale lightweight hash set is cross-scale fused and compensated to output a fusion horizontal compensation value. After the multi-scale sub-codon set is cross-scale aligned with the multi-scale lightweight hash set, a cross-scale parity sequence lightweight hash right shift accumulation is performed, and an iteration output vertical parity sequence accumulation result is output. The vertical parity sequence accumulation result is arithmetically summed to output an aggregated vertical compensation value. The fusion horizontal compensation value and the aggregated vertical compensation value are bitwise XORed to output the synthetic hash.

[0041] Cross-scale fusion compensation is a process of adjusting and optimizing lightweight hashes from different scales. Since the multi-scale lightweight hash set may have different biases, distributions or features, simply adding them directly may not get the best results. Therefore, cross-scale fusion compensation can balance the information between different scales. For example, standardize each scale's lightweight hash to conform to the same numerical range, such as by dividing by the standard deviation or min-max normalization, so that each lightweight hash has a relatively balanced influence. By adjusting these lightweight hashes through cross-scale fusion compensation, the differences between scales are eliminated, allowing them to be compared or combined under the same standard. This ensures that lightweight hashes from different scales can more fairly participate in the calculation of the final result. After cross-scale fusion compensation, the output fusion horizontal compensation value is obtained, which represents the unified result after compensation between scales.

[0042] Cross-scale codon alignment refers to aligning the same position codons in different scales. Since each codon set has different structures in different scales, it is necessary to align these codon sets in position. After alignment, cross-scale synucleotide lightweight hash right shift accumulation is performed, which means moving and accumulating lightweight hashes in the same position between multiple scales. Right shift accumulation means shifting the lightweight hash of a certain scale to the right and accumulating it with the hash value of the corresponding position of other scales. Right shift is a position adjustment of lightweight hash through bit shift operation, so that lightweight hashes of multiple scales can be effectively combined under the same framework. Accumulation is a superposition operation on each lightweight hash to form a new and more comprehensive hash value. In this way, information from different scales can be effectively integrated, and the differences between scales can be gradually reduced.

[0043] This process is performed by multiple iterations of right shift and accumulation to gradually obtain more stable and fused vertical synucleotide accumulation results. Vertical synucleotide accumulation results refer to the final hash value obtained by multiple right shifts and accumulations of hash information from multiple scales at the same position. This result can better reflect the comprehensive characteristics of the password and improve the security of the password.

[0044] Arithmetic summation is an addition operation on the vertical synucleotide accumulation result. Specifically, the sum of each bit or each part of the hash value is calculated to obtain a new aggregation result, i.e., the aggregation vertical compensation value. Through arithmetic summation, the results of vertical accumulation can be further integrated, allowing lightweight hashes from different scales to be better integrated into a whole, enhancing the complexity and information quantity of the password.

[0045] The bitwise XOR operation refers to comparing the corresponding bits of two binary numbers. If the corresponding bits are the same, the result is 0; if they are different, the result is 1. For fusing the horizontal compensation value and the aggregated vertical compensation value, the final synthetic hash is obtained by the bitwise XOR operation. This synthetic hash combines hash information from multiple scales and different compensations, further increasing the security of the password.

[0046] Further, the multi-scale lightweight hash set is compensated by cross-scale fusion, and a fused horizontal compensation value is output. The method comprises:

[0047] Adjacent hash horizontal compensation of the first scale lightweight hash set is performed according to the order of the sub-passwords in the first scale sub-password set, and a first horizontal compensation value is output. By analogy, multi-scale horizontal compensation values of the multi-scale sub-password set are obtained. The multi-scale horizontal compensation values are fused by using a recursive parity sequence fusion strategy, and the fused horizontal compensation value is output.

[0048] The first scale sub-password set is any one of the multi-scale sub-password set, which is the current analysis object. In the first scale sub-password set, the order of the sub-passwords is determined according to the arrangement order of the sub-passwords in the first scale sub-password set. Each sub-password represents a part of the password and contains some characteristics of the password. Adjacent hash horizontal compensation refers to performing hash compensation on adjacent sub-passwords in the first scale sub-password set. The purpose of this compensation is to adjust the relationship between adjacent sub-passwords so that their hash values have better matching when fused. For example, by dividing by the standard deviation or the minimum maximum value normalization, the influence of each lightweight hash is relatively balanced. By performing horizontal compensation on the hash values of adjacent sub-passwords, the influence of differences between adjacent sub-passwords can be eliminated, ensuring their uniformity in subsequent processing. After adjacent hash horizontal compensation, the first horizontal compensation value is obtained, which is the hash compensation result of the first scale sub-password set. This compensation value provides a balanced and optimized basis for subsequent fusion operations.

[0049] Similar to the adjacent hash horizontal compensation process of the first scale sub-password set, the same operation is performed on all multi-scale sub-password sets to obtain multi-scale horizontal compensation values. In this way, the horizontal compensation value of each scale can effectively reflect the characteristics of the sub-password set of that scale and prepare for the next fusion step.

[0050] Recursive parity bit fusion refers to fusing multi-scale transverse compensation values through the way of parity bits. Under this strategy, first, the transverse compensation values of each scale are grouped according to the order, and the compensation values in odd positions and even positions are separated. Recursive means that the fusion operation is performed multiple times. In each iteration, the values in odd positions and even positions are weighted or operated differently, and gradually fused into a final fused transverse compensation value. The recursive parity bit fusion strategy can ensure that the transverse compensation values of each scale can be fused according to different weights and positions in multiple recursions, forming a more complex and more balanced final result.

[0051] Further, the weak password rule library is used to mark the hit matching of the weak password by using the lightweight hash, and a password state matching result is output. The method comprises the following steps:

[0052] The Bloom filter performs existence detection on the lightweight hash based on a pre-constructed weak password lightweight hash value index. If the Bloom filter returns nonexistence, the password state matching result is directly output, wherein the password state matching result is a miss. If the Bloom filter returns possible existence, the weak password rule library is triggered to perform fine-grained matching on the lightweight hash. When the fine-grained matching misses, keyboard path variation detection is triggered based on the initial registration password, and a variation registration password is calculated and output. The variation hash of the variation registration password is used to match the weak password rule library to output the password state matching result.

[0053] The Bloom filter is a probabilistic data structure used to quickly detect whether an element exists in a set. Its advantages are small space occupation and fast lookup speed, but it may have a false negative possibility, that is, the Bloom filter may incorrectly report that an element exists, but it will not incorrectly report that it does not exist. In the system, a weak password lightweight hash value index has been pre-constructed. This index contains the hash values of common weak passwords, including known simple passwords, dictionary passwords or easily guessed passwords. The Bloom filter performs existence detection on the lightweight hash based on the pre-constructed weak password lightweight hash value index, that is, the lightweight hash of the password submitted by the user is input into the Bloom filter, and then the Bloom filter checks whether the lightweight hash exists in the weak password set.

[0054] If the Bloom filter detects that the lightweight hash of the submitted password does not exist in the weak password lightweight hash value index, that is, it returns nonexistence, it indicates that the password is not a weak password, and the subsequent operation can continue. In this case, the password state matching result is directly output as a miss.

[0055] If the Bloom filter returns a possible match, it indicates that the lightweight hash of the password might belong to a weak password, i.e. within the detection range of the Bloom filter. Since the Bloom filter is probabilistic, it might falsely report some elements not in the set, in this case, further trigger more accurate password checking. The weak password rule base is a database containing weak password rules, which stores various passwords that meet the weak password criteria, and the fine-grained matching refers to more detailed matching checks on the password, such as checking whether the password contains common dictionary words, uses simple number and letter combinations, or has known weaknesses. Through fine-grained matching, it can further confirm whether there is a weak password, even if the Bloom filter reports a possible match, this fine-grained matching step can effectively reduce false positives and provide users with more accurate password security assessment.

[0056] After fine-grained matching, if the password is still not determined as a weak password, it means that the password does not directly match the weak password in the weak password rule base, but in order to further enhance the complexity and security of the password, keyboard path variation detection is further performed at this link. Keyboard path variation refers to some changes that simulate user errors when inputting passwords due to physical or habitual errors, for example, the user may input incorrect characters on the keyboard, or intentionally insert adjacent characters in the password to increase password strength, for example, the letter a is mistakenly input as s. Variation detection identifies the possibility of these variations based on the initial registered password based on the physical path on the keyboard, generates a variation registered password, and the generation of the variation registered password changes the original password, further increasing the diversity and unpredictability of the password.

[0057] For the generated variation registered password, calculate its variation hash, which is similar to the original lightweight hash, but it is calculated for the variation password, and use the variation hash to match in the weak password rule base to see if the variation registered password belongs to a known weak password. If the variation hash finds a match in the weak password rule base, the password status matching result is output as a hit, indicating that the variation registered password is a weak password; if no match is found, the password status matching result is output as a miss, indicating that the variation registered password does not meet the weak password rule.

[0058] Further, the parallel feature vector is loaded into a pre-constructed multi-strategy fusion model, and multi-strategy password strength characteristic fusion evaluation is performed via the multi-strategy fusion model to output a real-time password strength label, and the method comprises:

[0059] Decompose the parallel feature vector to separate vector structure features, vector full-dimension data, and vector cryptography features; map the vector structure features, vector full-dimension data, and vector cryptography features to the rule engine channel, machine learning channel, and policy adaptation channel of the multi-strategy fusion model in parallel, analyze and output rule scores, risk probability values, and policy levels; dynamically weight the rule scores, risk probability values, and policy levels to generate a password strength score; and use the password strength score for label mapping to obtain the real-time password strength label.

[0060] The parallel feature vector is a plurality of password feature representations of the initial registered password, including vector structure features, vector full-dimension data, and vector cryptography features. The vector structure features refer to the composition and form of the password, such as whether the password contains letters, numbers, symbols, etc., as well as the length of the password, the distribution of characters, the arrangement between characters, etc. These vector structure features help to evaluate the complexity of the password and the degree of compliance with common password construction rules. The vector full-dimension data refers to the comprehensive features of the password, covering all information and attributes in the password. It is information extracted from all character dimensions of the password or other metadata contained in the password generation process. The vector cryptography features refer to information related to the cryptographic properties of the password. The security of the password depends not only on its surface structure, but also on its encryption strength and resistance to cracking. The vector cryptography features include whether the password is susceptible to dictionary attacks, brute-force attacks, or other cracking techniques, or whether the password meets the standards of certain encryption algorithms, etc.

[0061] Mapping the vector structure features, vector full-dimension data, and vector cryptography features to different analysis channels of the multi-strategy fusion model, the rule engine channel, machine learning channel, and policy adaptation channel are three parallel channels in the multi-strategy fusion model. They handle different types of features respectively. The rule engine channel processes the input feature data according to a predefined rule set, evaluating the compliance and complexity of the password. The machine learning channel analyzes the password features through a trained machine learning model, evaluating the potential risk or strength of the password. The policy adaptation channel adapts the input feature data according to specific application scenarios and policies, optimizing the password security evaluation.

[0062] After processing by each channel, the rule score, risk probability value, and policy level are output. The rule score is based on the evaluation results of the password features by the rule engine, indicating whether the password meets certain predefined rules, such as whether it contains common weak password patterns. The risk probability value is the output result of the machine learning channel, indicating the potential risk of the password. It is usually a numerical value representing the probability of the password being attacked or cracked. The policy level is the output result of the policy adaptation channel, indicating the security level of the password under a specific policy.

[0063] According to the predefined rules or real-time context, weights are dynamically assigned to rule scores, risk probability values, and policy levels, and each evaluation result has different importance in different application scenarios. For example, in a high security requirement scenario, a higher weight is given to the risk probability because the risk of cracking is more important. Through dynamic weighting fusion, the three evaluation results are weighted and calculated according to their weights, and finally a comprehensive password strength score is obtained. This score is a comprehensive evaluation of password security, which integrates information from multiple angles.

[0064] According to the password strength score, it is converted into a real-time password strength label, which is usually visualized and easy to understand. For example: the password strength score from 0 to 0.7 is a weak password, indicating that the password is less complex and is easy to crack; the password strength score from 0.7 to 1 is a strong password, indicating that the password has high complexity and strength and can effectively resist cracking attacks. This real-time password strength label reflects the security of the current password and can be updated in real time when the user modifies the password or submits the password.

[0065] Further, according to the real-time password strength label matching asynchronous scenarios, the initial registration password is subjected to asynchronous double-thread application scenario risk detection, and a heterogeneous scenario risk report is output. The method comprises:

[0066] If the real-time password strength label is a weak label type, a double-thread detection scenario is triggered, wherein the double-thread detection scenario includes historical password association detection and password leakage state detection; a user ID and the initial registration password input thread A are locally called to execute the historical password association detection, and output historical password repetition frequency, password minimum edit distance, and historical password similarity; the historical password repetition frequency, password minimum edit distance, and historical password similarity are dynamically weighted and quantified to output a password risk score; after generating a K-anonymous token based on the initial registration password, the K-anonymous token input thread B is executed to perform the password leakage state detection, and a password exposure frequency is output; the password risk score, password exposure frequency, and real-time password strength label are integrated to output a JSON structured heterogeneous scenario risk report.

[0067] If the real-time password strength label is a weak label type, i.e. the strength of the password is low, the weak password is easy to be guessed or cracked by common attack methods, and therefore needs more in-depth security check. At this time, the dual-thread detection scenario is started, which means that two main detection tasks, i.e. historical password correlation detection and password leakage state detection, are carried out in parallel during the password detection process, wherein the historical password correlation detection is to check the historical password record of the user to evaluate the similarity and potential risk of the current password with the historical password; the password leakage state detection is to check whether the password has been leaked to an external data leakage database or an attacker database. Through the dual-thread mode, multiple detection tasks can be processed simultaneously and in parallel, improving the processing efficiency and shortening the detection time.

[0068] In thread A, the user ID and the initial registration password are called as input parameters, and the historical password correlation detection is performed, the user ID represents the user who is currently performing the password detection, and the initial registration password is the current password submitted by the user. The historical password correlation detection aims to check whether the current password of the user has too much similarity with the historical password of the user, so as to identify whether the user is constantly using similar weak passwords, which may lead to reduced password security.

[0069] Specifically, the historical password repository of the user is queried, which is usually several password records of the user in the past period of time, in this example, the last 3 times, the password used by the user in the past period of time is analyzed, and the current password is compared with the historical password to obtain the historical password repetition frequency, which represents the repetition degree of the password used by the user before, if the password repetition frequency is high, the security of the password is low; the password minimum edit distance refers to the minimum number of operations, such as insertion, deletion and replacement of characters, required to convert one password into another password, which measures the difference between two passwords, for example, to convert the password password123 into passw0rd123, only one character needs to be replaced, and the minimum edit distance of this transformation is 1, which is used to measure the difference between the current password and the historical password, the smaller the distance, the worse the security; the historical password similarity is calculated by the password minimum edit distance, and the quotient value of the password minimum edit distance and the password length is calculated, and then 1 is subtracted from the quotient value to obtain the historical password similarity, which represents that if the edit distance between two passwords is small, their similarity will be high, and vice versa. The value is between 0 and 1, 1 means exactly the same, and 0 means completely different.

[0070] Dynamic weighting means dynamically adjusting the importance of each indicator according to specific security policies or scenarios, for example, if the historical password repetition frequency is high, give this indicator a higher weight, because the same password or similar password will increase the risk of being guessed or cracked, the quantitative process calculates a comprehensive password risk score according to the actual value of each indicator and its weight, the range of password risk score is 0 to 100, the higher the score, the higher the risk of password, the lower the security.

[0071] K-anonymous token is a privacy protection technology, by generating K-anonymous token, the initial registration password of the user is anonymized, K-anonymous is a data de-identification method, by processing the password, the same group of passwords cannot be associated with a specific user, thereby improving privacy protection. According to the generated K-anonymous token, the password leakage state detection is executed through thread B, specifically, querying the database or security service to detect whether the password has ever appeared in known password leakage events, for example, the password may have been exposed in a data leakage event and obtained by an attacker, the password exposure frequency refers to the number of times the password is leaked or the frequency of appearance in the leakage event, this indicator helps to assess whether the password has been widely exposed, further affecting its security, if the password appears frequently in the leakage event, it means that the password has a higher risk.

[0072] Integrate the obtained password risk score, password exposure frequency and real-time password strength label, these information combined together can provide a comprehensive perspective for analyzing the potential security problems of password. According to the comprehensive evaluation result, generate a JSON structured heterogeneous scene risk report, the JSON format output is convenient for system integration with other application programs, and can be easily provided to users, administrators or security team for further analysis and processing.

[0073] Further, if the real-time password strength label is a strong label type, a single-thread detection scenario is triggered, wherein the detection strategy of the single-thread detection scenario is password leakage state detection.

[0074] If the real-time password strength label is a strong label type, it indicates that the password has high complexity and strong anti-cracking ability, in this case, although the password is rated as a strong password, it is still necessary to confirm that the password has not been leaked to the outside, if the password has been leaked, its strength cannot effectively protect the security of the user's account. Therefore, the single-thread detection scenario is triggered, focusing on password leakage status detection, that is, checking whether the password has appeared in a known password leakage database, single-thread detection means that password leakage detection is independently executed without parallel with other tasks, and password leakage status detection checks whether the password has appeared in a past leakage event by comparing the password or related information input by the user with an external leakage data set, if the password appears in the known leakage database, the password is marked as having a leakage risk, even if it is a strong password. A leaked password will significantly reduce the security of the account when it is used, therefore, even if the password is very strong, its effectiveness will be reduced if it has been leaked.

[0075] In summary, the multi-strategy fusion password detection method provided by the embodiments of the present application has the following technical effects:

[0076] By calculating the lightweight hash of the initial registration password, the processing efficiency of password verification can be greatly improved. The lightweight hash algorithm is faster and consumes less resources than the traditional hash algorithm, reducing the load of the system while ensuring the security of password processing. This optimization enables the password detection to respond quickly and handle large-scale user-submitted passwords, improving the real-time performance and response speed of the system. By using a preloaded weak password rule library to quickly match the lightweight hash, weak passwords can be efficiently identified and intercepted. This matching process not only effectively prevents users from using easily cracked passwords, but also ensures the efficiency and accuracy of password detection, further improving detection efficiency. By using the SIMD (Single Instruction Multiple Data) instruction set for parallel processing of password feature extraction, the complex feature extraction process of the password can be accelerated. The SIMD instruction set allows the computer to process multiple data elements simultaneously in one operation, significantly improving processing speed. This technology improves the performance and efficiency of password detection, ensuring that the system can quickly process user-submitted passwords. The multi-strategy fusion model combines different evaluation strategies to more accurately and comprehensively evaluate the strength of the password, enhancing the intelligence and diversity of password detection. The multi-strategy fusion model outputs real-time password strength labels, enabling real-time response to user input and improving user experience. By using asynchronous dual-thread application scenario risk detection, multiple password risk detection tasks can be executed simultaneously, which not only improves detection efficiency but also avoids performance bottlenecks caused by multi-task execution. The application of the dual-thread mode enables the system to complete comprehensive evaluation of the password in the shortest time, providing timely security feedback to the user. By randomly perturbing the initial registration password to generate a temporary backup password library, the diversity of the password can be further enhanced, avoiding the use of simple or known password combinations. This provides an intelligent way for the system to deal with weak password selection by users, effectively increasing the complexity and unpredictability of the password and improving account security. Loading the temporary backup password library into the registration interface for visual display provides an interactive way for users to view and select passwords. This display method helps users easily select stronger passwords, further enhancing password security management.

[0077] In the second embodiment, based on the same inventive concept as the password detection method of the multi-strategy fusion in the preceding embodiments, as shown in the accompanying drawings, the present application provides a multi-strategy fusion password detection system, which comprises: Figure 2

[0078] A lightweight hash calculation module 10 is used to calculate the lightweight hash of the initial registration password after receiving the initial registration password submitted by the user.

[0079] ​The label hit matching module 20 is configured to perform weak password label hit matching on a weak password rule base using the lightweight hash, and output a password state matching result. The weak password rule base is preloaded when the user IU interface is switched to the registration interface.

[0080] The parallel feature vector extraction module 30 is configured to extract a parallel feature vector of the initial registration password via a SIMD instruction set if the password state matching result is a miss.

[0081] The strength characteristic fusion evaluation module 40 is configured to load the parallel feature vector into a pre-constructed multi-strategy fusion model, perform multi-strategy password strength characteristic fusion evaluation via the multi-strategy fusion model, and output a real-time password strength label.

[0082] The application scenario risk detection module 50 is configured to perform asynchronous dual-thread application scenario risk detection on the initial registration password according to the real-time password strength label matching an asynchronous scenario, and output a heterogeneous scenario risk report.

[0083] The random perturbation module 60 is configured to randomly perturb the initial registration password according to the heterogeneous scenario risk report to obtain a temporary backup password library.

[0084] The backup password library loading module 70 is configured to load the temporary backup password library to the registration interface for visual display.

[0085] Further, the lightweight hash calculation module 10 is configured to perform the following operation steps:

[0086] The user registration information is extracted on the registration interface, and a multi-element associated string is constructed based on a preset rule. The initial registration password is expanded to obtain an expanded registration password. The expanded registration password is subjected to multi-scale segmentation to obtain a multi-scale sub-password set. After calculating a multi-scale lightweight hash set of the multi-scale sub-password set, the multi-scale lightweight hash set is accumulated across scales to output a synthetic hash. After calculating an initial hash of the initial registration password, the synthetic hash and the initial hash are weighted based on a preset weight distribution to output the lightweight hash.

[0087] Further, the lightweight hash calculation module 10 is configured to perform the following operation steps:

[0088] The multi-scale lightweight hash set is compensated by cross-scale fusion, and a fusion horizontal compensation value is output. After the multi-scale sub-cipher set is aligned by cross-scale parity sub-cipher, cross-scale parity sequence lightweight hash right shift accumulation of the multi-scale lightweight hash set is performed, and a vertical parity sequence accumulation result is iteratively output. The vertical parity sequence accumulation result is arithmetically summed, and an aggregated vertical compensation value is output. The fusion horizontal compensation value and the aggregated vertical compensation value are subjected to a bitwise XOR operation, and the synthetic hash is output.

[0089] Further, the lightweight hash calculation module 10 is configured to perform the following operation steps:

[0090] Adjacent hash horizontal compensation of the first scale lightweight hash set is performed according to the sub-cipher order in the first scale sub-cipher set, and a first horizontal compensation value is output. The multi-scale horizontal compensation values of the multi-scale sub-cipher set are obtained by analogy. The multi-scale horizontal compensation values are fused by using a recursive parity sequence bit fusion strategy, and the fusion horizontal compensation value is output.

[0091] Further, the tag hit matching module 20 is configured to perform the following operation steps:

[0092] The Bloom filter performs existence detection on the lightweight hash based on a pre-constructed weak password lightweight hash value index. If the Bloom filter returns non-existence, the password state matching result is directly output, wherein the password state matching result is a miss. If the Bloom filter returns possible existence, the weak password rule library is triggered to perform fine-grained matching on the lightweight hash. When fine-grained matching misses, keyboard path variation detection is triggered based on the initial registration password, and a variation registration password is calculated and output. The variation registration password is used to perform variation hash matching on the weak password rule library to output the password state matching result.

[0093] Further, the strength characteristic fusion evaluation module 40 is configured to perform the following operation steps:

[0094] The parallel feature vector is decomposed to separate vector structure features, vector full-dimension data, and vector cryptography features. The vector structure features, vector full-dimension data, and vector cryptography features are mapped to the rule engine channel, machine learning channel, and strategy adaptation channel in the multi-strategy fusion model in parallel, and rule scores, risk probability values, and strategy levels are analyzed and output. The rule scores, risk probability values, and strategy levels are dynamically weighted and fused to generate a password strength score. The password strength score is used for label mapping to obtain the real-time password strength label.

[0095] Further, the application scenario risk detection module 50 is configured to perform the following operation steps:

[0096] If the real-time password strength label is a weak label type, a double-thread detection scenario is triggered, wherein the double-thread detection scenario comprises historical password association detection and password leakage state detection; a user ID and the initial registration password input thread A are locally called to perform the historical password association detection, and output historical password repetition frequency, password minimum edit distance and historical password similarity; the historical password repetition frequency, password minimum edit distance and historical password similarity are quantified by dynamic weighting, and a password risk score is output; after a K-anonymous token is generated based on the initial registration password, the K-anonymous token is input into a thread B to perform the password leakage state detection, and a password exposure frequency is output; the password risk score, the password exposure frequency and the real-time password strength label are integrated, and a JSON structured heterogeneous scenario risk report is output.

[0097] Further, if the real-time password strength label is a strong label type, a single-thread detection scenario is triggered, wherein the detection strategy of the single-thread detection scenario is password leakage state detection.

[0098] Through the foregoing detailed description of the password detection method of the multi-strategy fusion, those skilled in the art can clearly understand the password detection system of the multi-strategy fusion in the embodiment. Since the system corresponds to the method disclosed in the embodiment, the system is described relatively simply, and the relevant part can be referred to the method part.

[0099] Embodiment three provides a storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement any step of the method of embodiment one.

[0100] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present disclosure.

[0101] The above description of disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-strategy fusion password detection method, characterized in that, The method comprises: After receiving the initial registration password submitted by the user, calculating the lightweight hash of the initial registration password; Weak password rule library is used for weak password marking hit matching, and the password state matching result is output, wherein the weak password rule library is preloaded when the user IU interface is switched to the registration interface; If the password state matching result is not hit, the parallel feature vector of the initial registration password is extracted through the SIMD instruction set; Load the parallel feature vector into the pre-constructed multi-strategy fusion model, perform multi-strategy password strength characteristic fusion evaluation through the multi-strategy fusion model, and output real-time password strength label; According to the real-time password strength label matching asynchronous scene, the asynchronous double-thread application scene risk detection is carried out on the initial registration password, and the heterogeneous scene risk report is output; If the real-time password strength label is a weak label type, trigger a double-thread detection scene, wherein the double-thread detection scene includes historical password association detection and password leakage state detection; If the real-time password strength label is a strong label type, trigger a single-thread detection scene, wherein the detection strategy of the single-thread detection scene is password leakage state detection; According to the heterogeneous scene risk report, the initial registration password is randomly disturbed to obtain a temporary backup password library; Load the temporary backup password library into the registration interface for visual display.

2. The multi-strategy fused password detection method of claim 1, wherein, After receiving the initial registration password submitted by the user, calculating the lightweight hash of the initial registration password, the method comprises: Extracting user registration information on the registration interface, and constructing a multi-element associated string based on a preset rule; The initial registration password is expanded by using the multi-element associated string to obtain an expanded registration password; The expanded registration password is subjected to multi-scale segmentation to obtain a multi-scale sub-password set; After calculating the multi-scale lightweight hash set of the multi-scale sub-password set, the multi-scale lightweight hash set is accumulated across the scale to output a synthetic hash; After calculating the initial hash of the initial registration password, the synthetic hash and the initial hash are weighted based on a preset weight distribution, and the lightweight hash is output.

3. The multi-strategy fused password detection method of claim 2, wherein, After calculating the multi-scale lightweight hash set of the multi-scale sub-password set, the multi-scale lightweight hash set is accumulated across the scale to output a synthetic hash, the method comprises: Cross-scale fusion compensation is performed on the multi-scale lightweight hash set to output a fusion horizontal compensation value; After aligning the multi-scale sub-password set across the scale, the cross-scale parity sequence lightweight hash right shift accumulation of the multi-scale lightweight hash set is mapped, until the iteration output vertical parity sequence accumulation result is output; The vertical parity sequence accumulation result is subjected to arithmetic summation to output an aggregated vertical compensation value; The fusion horizontal compensation value and the aggregated vertical compensation value are subjected to bitwise XOR operation to output the synthetic hash.

4. The multi-strategy fused password detection method of claim 3, wherein, The method comprises: According to the order of the sub-password in the first scale sub-password set, the adjacent hash horizontal compensation of the first scale lightweight hash set is performed to output a first horizontal compensation value; By analogy, the multi-scale horizontal compensation values of the multi-scale sub-password set are obtained; The multiple-scale lateral compensation values are fused by using a recursive parity sequence bit fusion strategy to output the fused lateral compensation value.

5. The multi-strategy fused password detection method of claim 1, wherein, The lightweight hash is used for weak password marking hit matching in a weak password rule library to output a password state matching result, and the method comprises: A Bloom filter performs existence detection on the lightweight hash based on a pre-constructed weak password lightweight hash value index; If the Bloom filter returns non-existence, the password state matching result is directly outputted, wherein the password state matching result is a miss; If the Bloom filter returns possible existence, the weak password rule library is triggered to perform fine-grained matching on the lightweight hash; When fine-grained matching misses, keyboard path variation detection is triggered based on the initial registration password to calculate and output a variation registration password; The variation hash of the variation registration password is used for matching in the weak password rule library to output the password state matching result.

6. The multi-strategy fused password detection method of claim 1, wherein, The parallel feature vectors are loaded into a pre-constructed multi-strategy fusion model, multi-strategy password strength feature fusion evaluation is performed via the multi-strategy fusion model, and a real-time password strength label is outputted, and the method comprises: The parallel feature vectors are decomposed to separate vector structure features, vector full-dimension data and vector cryptography features; The vector structure features, vector full-dimension data and vector cryptography features are mapped synchronously into parallel rule engine channels, machine learning channels and strategy adaptation channels in the multi-strategy fusion model, and rule scores, risk probability values and strategy levels are analyzed and outputted; The rule scores, risk probability values and strategy levels are dynamically weighted and fused to generate a password strength score; The password strength score is used for label mapping to obtain the real-time password strength label.

7. The multi-strategy fused password detection method of claim 1, wherein, The triggered double-thread detection scenario specifically comprises: A local call user ID and the initial registration password input thread A perform the historical password association detection to output historical password repetition frequency, password minimum edit distance and historical password similarity; The historical password repetition frequency, password minimum edit distance and historical password similarity are dynamically weighted and quantified to output a password risk score; After a K-anonymity token is generated based on the initial registration password, the K-anonymity token is inputted into thread B to perform the password leakage state detection to output a password exposure frequency; The password risk score, password exposure frequency and real-time password strength label are integrated to output a JSON-structured heterogeneous scene risk report.

8. A multi-strategy fused password detection system, characterized in that, A multi-strategy fusion password detection method for implementing any one of claims 1-7, the system comprises: A lightweight hash calculation module configured to calculate a lightweight hash of an initial registration password after receiving the initial registration password submitted by a user; A marking hit matching module configured to perform weak password marking hit matching in a weak password rule library by using the lightweight hash to output a password state matching result, wherein the weak password rule library is preloaded when a user IU interface is switched to a registration interface; A parallel feature vector extraction module configured to extract a parallel feature vector of the initial registration password by using a SIMD instruction set if the password state matching result is a miss; and A multi-strategy fusion model configured to load the parallel feature vector into the multi-strategy fusion model, perform multi-strategy password strength feature fusion evaluation via the multi-strategy fusion model, and output a real-time password strength label. a strength property fusion evaluation module, configured to load the parallel feature vectors to a pre-constructed multi-strategy fusion model, perform multi-strategy password strength property fusion evaluation via the multi-strategy fusion model, and output a real-time password strength label; an application scenario risk detection module, configured to match an asynchronous scenario according to the real-time password strength label, perform asynchronous double-thread application scenario risk detection on the initial registration password, and output a heterogeneous scenario risk report; a random perturbation module, configured to randomly perturb the initial registration password according to the heterogeneous scenario risk report, and obtain a temporary backup password library; a backup password library loading module, configured to load the temporary backup password library to a registration interface for visual display.

9. A storage medium having stored thereon a computer program, characterized in that The computer program, when executed by a processor, implements the steps of the multi-strategy fusion password detection method of any one of claims 1 to 7.

Images