Identity authentication and access control system based on security management
Through the deep learning model, the semantic understanding of user access factor data is solved, and the lack of unified management and secure access control in the business system is improved, the rationality and necessity of secondary strong authentication is improved, and the user experience and system security are improved.
Patent Information
- Application Number
- CN202310416380.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-04-18
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2043-04-18
AI Technical Summary
The multiple independent constructions of existing business systems have led to a lack of unified management and secure access control, which poses major management problems and security risks. The rationality and necessity of secondary strong authentication are difficult to judge, affecting user experience and security.
The semantic encoder containing the embedded layer and the multi-scale neighborhood feature extraction module are used to semantically understand the user's access factor data through a deep learning model, generate classification results for secondary strong authentication, and improve their rationality and necessity.
It improves the rationality and necessity of secondary strong authentication, improves user experience, and at the same time enhances the security and management efficiency of the business system.
Smart Images

Figure CN116522197B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of security protection, and more specifically to an identity authentication and access control system based on security management. Background Art
[0002] With the continuous deepening of informatization construction, business systems have brought qualitative changes to the work efficiency and management of enterprises and institutions. Business systems have become the core assets of various enterprises and institutions, and are directly related to the healthy development and economic interests of enterprises and institutions. The security protection construction of business systems has become the top priority of information security construction.
[0003] Since the construction time, construction departments, and construction manufacturers of multiple business systems of various enterprises and institutions are relatively independent, and the main construction requirements of most business systems focus more on business logic processing, there is a lack of unified management of multiple business systems and security access control considerations, which leads to major management problems and security risks.
[0004] Therefore, an optimized identity authentication and access control system based on security management is expected. Summary of the invention
[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an identity authentication and access control system based on security management, which takes into account that the access factor data of the detected user implies the access behavior characteristics of the detected user. Therefore, a semantic encoder including an embedding layer is used to mine the global implicit associations between the various access factor data to obtain multiple contextual access factor semantic vectors, and further, a multi-scale understanding is performed on the obtained multiple contextual access factor semantic vectors. In this way, the access behavior semantics of the access factor data of the user to be detected are understood based on the deep network model, and whether to generate a secondary strong authentication is determined based on the extracted semantic features, so as to improve the rationality and necessity of the generation of the secondary strong authentication.
[0006] According to one aspect of the present application, there is provided an identity authentication and access control system based on security management, which includes:
[0007] An access factor data monitoring unit is used to obtain access factor data of the user to be detected, wherein the access factor data includes: user access time, location, sensitivity of the business system and access behavior;
[0008] An access factor semantic understanding unit, used for passing the access factor data of the user to be detected through a semantic encoder including an embedding layer to obtain a plurality of contextual access factor semantic vectors;
[0009] A cascading unit, configured to cascade the plurality of context access element semantic vectors to obtain a global access semantic feature vector;
[0010] A multi-scale semantic association feature extraction unit, used for one-dimensionally arranging the plurality of context access element semantic vectors into feature vectors and then passing them through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector;
[0011] a multi-scale fusion unit, configured to fuse the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain a classification feature vector; and
[0012] The access control generation unit is used to pass the classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether to generate a secondary strong authentication.
[0013] In the above-mentioned identity authentication and access control system based on security management, the access factor semantic understanding unit includes: an embedding unit, which is used to use the embedding layer of the context encoder to map each data item in the access factor data of the user to be detected into an embedding vector to obtain a sequence of embedded vectors; and a context encoding unit, which is used to use the context encoder to perform converter-based context semantic encoding on the sequence of embedded vectors to obtain the multiple context access factor semantic vectors.
[0014] In the above-mentioned identity authentication and access control system based on security management, the context encoding unit is further used to: arrange the sequence of the embedded vectors into an input vector; convert the input vector into a query vector and a key vector respectively through a learnable embedding matrix; calculate the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; normalize the self-attention association matrix to obtain a standardized self-attention association matrix; input the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and multiply the self-attention feature matrix with each embedded vector in the sequence of the embedded vectors as a value vector to obtain the multiple context access element semantic vectors.
[0015] In the above-mentioned identity authentication and access control system based on security management, the cascading unit is further used to: cascade the multiple context access element semantic vectors to obtain a global access semantic feature vector according to the following formula; wherein the formula is:
[0016] X=Concat[V1,V2,...,V k ]
[0017] Among them, V1, V2, ..., V k represents the semantic vector of each context access element, and Concat[·,·,·] represents the cascade function.
[0018] In the above-mentioned identity authentication and access control system based on security management, the multi-scale semantic association feature extraction unit includes: a first-scale semantic association feature extraction subunit, used to input the feature vector into the first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first-scale neighborhood access semantic feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel of a first length; a second-scale semantic association feature extraction subunit, used to input the feature vector into the second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second-scale neighborhood access semantic feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and a multi-scale cascade subunit, used to cascade the first-scale neighborhood access semantic feature vector and the second-scale neighborhood access semantic feature vector to obtain the multi-scale neighborhood access semantic feature vector.
[0019] In the above-mentioned identity authentication and access control system based on security management, the first-scale semantic association feature extraction subunit is further used to: use the first convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the feature vector according to the following formula to obtain the first-scale neighborhood access semantic feature vector; wherein the formula is:
[0020]
[0021] Wherein, a is the width of the first convolution kernel in the x direction, F(a) is the first convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the feature vector; the second-scale semantic association feature extraction subunit is further used to: use the second convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the feature vector according to the following formula to obtain the second-scale neighborhood access semantic feature vector; wherein the formula is:
[0022]
[0023] Among them, b is the width of the second convolution kernel in the x direction, F(b) is the second convolution kernel parameter vector, G(xb) is the local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, and X represents the eigenvector.
[0024] In the above-mentioned identity authentication and access control system based on security management, the multi-scale fusion unit is further used to calculate the position-weighted sum between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain the classification feature vector.
[0025] In the above-mentioned identity authentication and access control system based on security management, the access control generation unit is further used to: use the classifier to process the classification feature vector according to the following formula to obtain the classification result, wherein the formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is the bias vector, and X is the classification feature vector.
[0026] In the above-mentioned identity authentication and access control system based on security management, it also includes a training module for training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier; wherein the training module includes: a training data acquisition unit, used to acquire training data, the training data includes access factor data of the training user to be detected, and a true value of whether the training user to be detected generates a secondary strong authentication; a training access factor semantic understanding unit, used to pass the access factor data of the training user to be detected through the semantic encoder including the embedding layer to obtain multiple training context access factor semantic vectors; a training cascade unit, used to cascade the multiple training context access factor semantic vectors to obtain a training global access semantic feature vector; a training multi-scale semantic association feature extraction unit, used to cascade the multiple training context access factor semantic vectors to obtain a training global access semantic feature vector. The semantic vector is arranged in one dimension into a training feature vector and then passed through the multi-scale neighborhood feature extraction module to obtain a training multi-scale neighborhood access semantic feature vector; a training multi-scale fusion unit is used to fuse the training global access semantic feature vector and the training multi-scale neighborhood access semantic feature vector to obtain a training classification feature vector; a classification loss unit is used to pass the training classification feature vector through the classifier to obtain a classification loss function value; an internalized learning loss unit is used to calculate the sequence-to-sequence response rule internalized learning loss function value between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector; and a training unit is used to train the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier using the weighted sum of the internalized learning loss function value and the classification loss function value as the loss function value pair.
[0027] In the above-mentioned identity authentication and access control system based on security management, the internalized learning loss unit is further used to calculate the sequence-to-sequence response rule internalized learning loss function between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector using the following formula; wherein the formula is:
[0028]
[0029]
[0030]
[0031] Among them, V1 and V2 are the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector respectively, and W1 and W2 are the weight matrices of the classifier for V1 and V2 respectively.
[0032] According to another aspect of the present application, there is also provided a method for identity authentication and access control based on security management, which includes:
[0033] Obtaining access factor data of the user to be detected, wherein the access factor data includes: user access time, location, sensitivity of the business system and access behavior;
[0034] Passing the access factor data of the user to be detected through a semantic encoder including an embedding layer to obtain a plurality of contextual access factor semantic vectors;
[0035] Cascading the multiple context access element semantic vectors to obtain a global access semantic feature vector;
[0036] Arranging the multiple context access element semantic vectors in one dimension into feature vectors and then passing them through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector;
[0037] fusing the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain a classification feature vector; and
[0038] The classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether to generate a secondary strong authentication.
[0039] In the above-mentioned identity authentication and access control method based on security management, the access factor data of the user to be detected is passed through a semantic encoder including an embedding layer to obtain multiple contextual access factor semantic vectors, including: using the embedding layer of the context encoder to map each data item in the access factor data of the user to be detected into an embedding vector to obtain a sequence of embedded vectors; and using the context encoder to perform converter-based contextual semantic encoding on the sequence of embedded vectors to obtain the multiple contextual access factor semantic vectors.
[0040] In the above-mentioned identity authentication and access control method based on security management, the context encoder is used to perform converter-based context semantic encoding on the sequence of embedded vectors to obtain the multiple context access element semantic vectors, including: arranging the sequence of embedded vectors as input vectors; converting the input vectors into query vectors and key vectors respectively through learnable embedding matrices; calculating the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; normalizing the self-attention association matrix to obtain a standardized self-attention association matrix; inputting the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and multiplying the self-attention feature matrix with each embedded vector in the sequence of embedded vectors as a value vector to obtain the multiple context access element semantic vectors.
[0041] In the above-mentioned identity authentication and access control method based on security management, it also includes training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier; wherein, the process of training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier includes the steps of: obtaining training data, the training data including access factor data of the training user to be detected, and a true value of whether the training user to be detected generates a secondary strong authentication; passing the access factor data of the training user to be detected through the semantic encoder including the embedding layer to obtain multiple training context access factor semantic vectors; cascading the multiple training context access factor semantic vectors to obtain a training global access semantic feature vector; and The training context access element semantic vector is arranged in one dimension into a training feature vector and then passed through the multi-scale neighborhood feature extraction module to obtain a training multi-scale neighborhood access semantic feature vector; the training global access semantic feature vector and the training multi-scale neighborhood access semantic feature vector are fused to obtain a training classification feature vector; the training classification feature vector is passed through the classifier to obtain a classification loss function value; the sequence-to-sequence response rule internalization learning loss function value between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector is calculated; and the weighted sum of the internalization learning loss function value and the classification loss function value is used as the loss function value pair to train the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier.
[0042] Compared with the prior art, the identity authentication and access control system based on security management provided by the present application takes into account that the access factor data of the detected user implies the access behavior characteristics of the detected user. Therefore, a semantic encoder including an embedding layer is used to mine the global implicit associations between the various access factor data to obtain multiple contextual access factor semantic vectors, and further, a multi-scale understanding is performed on the obtained multiple contextual access factor semantic vectors. In this way, the access behavior semantics of the access factor data of the detected user is understood based on the deep network model, and whether to generate a secondary strong authentication is determined based on the extracted semantic features, so as to improve the rationality and necessity of generating the secondary strong authentication. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0044] Figure 1 This is an application scenario diagram of the application security access control system according to an embodiment of the present application.
[0045] Figure 2 A block diagram of an identity authentication and access control system based on security management according to an embodiment of the present application.
[0046] Figure 3 The present invention is an architecture diagram of an identity authentication and access control system based on security management according to an embodiment of the present application.
[0047] Figure 4 It is a block diagram of a training module in an identity authentication and access control system based on security management according to an embodiment of the present application.
[0048] Figure 5 The present invention is a flowchart of an identity authentication and access control method based on security management according to an embodiment of the present application.
[0049] Figure 6 The present invention is a flowchart of training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier in the identity authentication and access control method based on security management according to an embodiment of the present application. DETAILED DESCRIPTION
[0050] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0051] Application Overview
[0052] As mentioned above, due to the relatively independent construction time, construction departments, and construction manufacturers of multiple business systems of various enterprises and institutions, and the main construction requirements of most business systems focus more on business logic processing, lack of unified management of multiple business systems and security access control considerations, there are major management problems and security risks. Therefore, an optimized identity authentication and access control system based on security management is expected.
[0053] Accordingly, in the technical solution of this application, an application security access control system is constructed to provide comprehensive security protection and user access control for the business system through terminal security containers and server security containers. The system adopts a structural model in which the data plane and the control plane are separated from each other, ensuring that a large amount of data processing does not affect management and control. In a complex network environment, the control plane and the management plane do not affect each other, which highly guarantees the security and stability of the system. In addition, based on factors such as user access time, location, sensitivity of the business system, and access behavior, users are required to undergo secondary strong authentication to ensure that the visitor is a legal and authorized natural person.
[0054] However, during the secondary strong authentication process, the applicant of this application found that if the secondary strong authentication is inappropriate, it will greatly affect the user experience, and if the secondary strong authentication is not timely, it will reduce the security of access. In view of the above technical problems, it is expected to be able to intelligently determine whether to generate a secondary strong authentication based on the access factor data of the detected user.
[0055] Specifically, the access factor data of the detected user implies the access behavior characteristics of the detected user, but this mapping relationship is implicit and nonlinear, and it is difficult to accurately mine and understand based on traditional machine learning or feature engineering methods. Accordingly, in recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in the fields of image classification, object detection, semantic segmentation, text translation, etc. The development of deep learning and neural networks provides new solutions and solutions for understanding user access behavior.
[0056] Correspondingly, in the technical solution of the present application, the access factor data of the user to be detected is first obtained, and the access factor data includes: user access time, location, sensitivity of the business system and access behavior. Then, the access factor data of the user to be detected is passed through a semantic encoder including an embedding layer to obtain a plurality of contextual access factor semantic vectors. Here, considering that each data in the access factor data is not structured data, therefore, before inputting the semantic understanding model for semantic understanding, each data item in the access factor data of the user to be detected is first structured. Specifically, the embedding layer of the context encoder is used to map each data item in the access factor data of the user to be detected into an embedding vector to obtain a sequence of embedded vectors, for example, it can be vector embedded through a learnable embedding matrix. Then, the context encoder is used to perform converter-based contextual semantic encoding on the sequence of embedded vectors to obtain the plurality of contextual access factor semantic vectors.
[0057] In order to fully mine the access behavior semantic information contained in the access factor data, the obtained multiple context access factor semantic vectors are further understood at a multi-scale. Specifically, the multiple context access factor semantic vectors are first cascaded to obtain a global access semantic feature vector. At the same time, the multiple context access factor semantic vectors are arranged in one dimension as feature vectors and then passed through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector. Here, the multi-scale neighborhood feature extraction module includes multiple parallel one-dimensional convolutional layers, which can use one-dimensional convolution kernels with different scales to perform multi-scale one-dimensional convolution encoding on the feature vector arranged by the multiple context access factor semantic vectors to extract the local neighborhood semantic association features between different access factors in the multiple context access factor semantic vectors to obtain the multi-scale neighborhood access semantic feature vector.
[0058] Next, the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector are fused to obtain a classification feature vector, and the obtained classification feature vector is classified by a classifier to obtain a classification result indicating whether to generate a secondary strong authentication. In this way, the access behavior semantics of the access element data of the user to be detected are understood based on the deep network model, and whether to generate a secondary strong authentication is determined based on the extracted semantic features, so as to improve the rationality and necessity of generating the secondary strong authentication and enhance the user experience.
[0059] Here, when the multi-scale neighborhood access semantic feature vector is obtained by a multi-scale neighborhood feature extraction module after the multiple context access element semantic vectors are arranged one-dimensionally into feature vectors, it is expected that the multi-scale neighborhood feature extraction module can extract the associated features of the context access element semantics under the multi-scale neighborhood while still having good responsiveness to the global access semantic feature vector obtained by directly cascading the multiple context access element semantic vectors, so as to enhance the fusion effect of the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector.
[0060] Therefore, the sequence-to-sequence response rule internalization learning loss function between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector is calculated:
[0061]
[0062]
[0063]
[0064] V1 and V2 are the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector, respectively, and M1 and M2 are the weight matrices of the classifier for V1 and V2, respectively.
[0065] That is, the enhanced distinguishing ability between sequences is obtained by using the squeeze-excite channel attention mechanism of the classifier for the weight matrices of different vector sequences represented by the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector. In this way, by training the network with this loss function, the recovery of causal relationship features with better distinguishing characteristics between response sequences can be achieved, so as to internalize the cause-effect response rules between sequences and improve the responsiveness of the multi-scale neighborhood access semantic feature vector to the global access semantic feature vector.
[0066] Based on this, the present application proposes an identity authentication and access control system based on security management, which includes: an access factor data monitoring unit, used to obtain the access factor data of the user to be detected, and the access factor data includes: user access time, location, sensitivity of the business system and access behavior; an access factor semantic understanding unit, used to pass the access factor data of the user to be detected through a semantic encoder including an embedding layer to obtain multiple context access factor semantic vectors; a cascade unit, used to cascade the multiple context access factor semantic vectors to obtain a global access semantic feature vector; a multi-scale semantic association feature extraction unit, used to arrange the multiple context access factor semantic vectors into feature vectors in one dimension and then pass them through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector; a multi-scale fusion unit, used to fuse the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain a classification feature vector; and an access control generation unit, used to pass the classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether to generate a secondary strong authentication.
[0067] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.
[0068] Exemplary Systems
[0069] With the continuous deepening of informatization construction, business systems have brought qualitative changes to the work efficiency and management of enterprises and institutions. Business systems have become the core assets of various enterprises and institutions, and are directly related to the healthy development and economic interests of enterprises and institutions. The security protection construction of business systems has become the top priority of information security construction.
[0070] Since the construction time, construction departments, and construction manufacturers of multiple business systems of various enterprises and institutions are relatively independent, and the main construction requirements of most business systems focus more on business logic processing, there is a lack of unified management of multiple business systems and security access control considerations, which leads to major management problems and security risks.
[0071] In terms of user management, multiple business systems are relatively independent, and the organizational structure and user management within each system are independent of each other. Changes such as employee recruitment, transfer, and resignation require independent operations on each business system, especially account cancellation or deletion operations after employee resignation, which are very easy to miss. It is not uncommon for employees to still be able to access the business system normally after resigning, posing a huge security risk.
[0072] In terms of identity authentication, most business systems still use static username and password authentication methods, which have low authentication strength, and there are security risks such as weak passwords and the same password for multiple systems, as well as the problem of the inability to uniformly manage password policies. If the authentication mechanism (MFA) of the business system needs to be upgraded or strengthened, there will be huge and uncontrollable transformation costs, and even the problem of old systems being unable to be transformed.
[0073] In terms of its own security, during the development / upgrade of business systems, there may be obvious security loopholes due to the lack of security awareness among developers. For example, some key interfaces are vulnerable to enumeration attacks because they have no verification or other prevention mechanisms. Some interfaces lack validation of user login credentials or have verification defects, allowing attackers to access sensitive information without authorization or even perform unauthorized operations.
[0074] In terms of authorized access, the authorization system of the business system is based on the functional authorization of platform roles. It is impossible to set flexible security access control policies based on dimensions such as user access time, access location, and user behavior. At the same time, the copying and downloading of sensitive data of the business system cannot be effectively controlled, and the security of the data cannot be effectively guaranteed. According to the Fortscal investigation report, 85% of leaks are caused internally.
[0075] In terms of comprehensive auditing, after the construction of independent business systems is completed, in addition to the above-mentioned difficulties, the problem of unified auditing has gradually become prominent. The location, content, and format of business system log records are not unified, which creates great technical obstacles for comprehensive auditing based on user access behavior.
[0076] Therefore, we launched the application security access control system to solve the above problems with a productized solution without the need to transform the business system.
[0077] The application security access control system performs comprehensive security protection and user access control on the business system through the terminal security container and the server security container. The system adopts a structural model in which the data plane and the control plane are separated from each other, ensuring that large amounts of data processing do not affect management and control. In a complex network environment, the control plane and the management plane do not affect each other, thereby highly ensuring the security and stability of the system.
[0078] Specifically, the application security access control system provides a terminal security container module, which can require user terminals to use the terminal security container to access the business system based on factors such as user, time, location, and business system: browser development and debugging mode control, terminal clipboard control, terminal operation recording, address bar locking, and input verification.
[0079] Specifically, the server-side security container is the core of the application security access control system. It logically protects the business system in the security container. All users' access to the business system must and can only be performed under the authentication, authorization, control and audit of the server-side security container. The business system is not visible to the outside world. Its functions include: secondary authentication, which can require users to perform secondary strong authentication based on factors such as user access time, location, sensitivity of the business system and access behavior to ensure that the visitor is a legal and authorized natural person; minimize authorization of internal system functions, and provide secondary authorization capabilities within the business system without any modification to the business system. It can deny the current user access to the core sensitive functions within the business system based on factors such as access time, access location, access behavior, etc. The control granularity is smaller than that of the business system control. Granularity; application approval, that is, providing workflow approval function. When the permissions required for user business development conflict with the security control policy, the user can apply for approval workflow and obtain relevant permissions temporarily with the permission of the approver, and access sensitive functions under control; page watermark, that is, providing business system page watermark function, and the business system does not need any modification. The page watermark display content information includes: user main account information, user secondary account information, access time, etc.; dynamic desensitization, that is, providing business system online data desensitization function; and upload and download approval, that is, providing business system file upload / download control function. The business system does not need any modification, and can prohibit uploading / downloading files or upload / download files through application approval, thereby improving system security and reducing data leakage risks.
[0080] Specifically, the application security access control system provides full life cycle management capabilities for user identities. It can achieve full life cycle management of employees' entry, transfer, and resignation through master-slave account management and master-slave account association. For those who are willing or have the integration capability, docking and integration can also be completed through interfaces.
[0081] Specifically, the application security access control system provides a unified portal for all business systems, and users can easily access all authorized business systems on a single page. Business personnel do not need to remember the IP ports / domain names of each business system and the corresponding usernames and passwords. The system has "anti-bypass" capabilities, and all users cannot directly establish connections with business systems.
[0082] Specifically, the application security access control system has a unified strong authentication capability. Authentication methods include username / password, certificate, OTP, SMS and other authentication methods. At the same time, it provides a flexible authentication policy management mechanism: user account / password life cycle, password complexity policy management; centralized strong authentication center, used to provide centralized authentication services for primary login authentication and secondary login authentication; authentication hub service, support various standard third-party authentication protocol integration, such as AD authentication, Radius authentication, etc.; any combination of multiple authentication methods, providing authentication means with different security levels and different security policies; and support for third-party interface authentication, which can be connected with third-party authentication according to the actual interface document.
[0083] Specifically, the application security access control system can realize the SSO (single sign-on) capability of the business system through standard interface docking or username and password filling. Users only need to log in once to access all authorized business systems. It also has a single sign-out capability. When a user logs out, all open business systems can be logged out at the same time.
[0084] Specifically, the application security access control system is based on the authorization model of the position, that is, through the position, the relationship between users and business systems is mapped (that is, who has the right to access what business system), and access to different business systems is authorized according to different user positions, so as to realize the centralized control of user rights. In addition, without any modification of the business system, the system provides secondary authorization capabilities within the business system, and can deny or apply for approval based on factors such as access time, access location, and access behavior, and control access to the core sensitive functions within the business system.
[0085] Specifically, the core component of the application security access control system: the server-side security container, which isolates the business system and business users securely and makes the business system "invisible". All access requests to the business system must be authenticated, authorized, controlled and audited by the application security access control system to prevent illegal scanning and illegal penetration of the business system by internal and external network personnel, and uniformly enhance the security of the business system. The core component of the application security access control system: the server-side security container, through deep security detection at the bottom layer of the protocol, can accurately match the attack characteristics, algorithms, and patterns of the application layer. At the same time, the application security access control system can also detect and alert abnormal business behaviors and operations. The core component of the application security access control system: the server-side security container has comprehensive security detection and intrusion prevention functions, which can effectively resist attack scanning and hacker penetration, and can also alert and block common attacks of OWASP Top10, such as WSQL injection, XSS and CSRF, and can resist CC and web slow attacks. At the same time, it can prevent Trojans, dark links and black links against the application system, and also provides page watermarking capabilities, which can effectively prevent phishing. In addition, the administrator provides remote monitoring methods to monitor the operating behaviors of key personnel on sensitive business systems in real time. Once violations or high-risk operations are found, the administrator can remotely lock / interrupt the business personnel's operation interface with one click, and can also send real-time alarm prompts. After the threat is lifted, it can be unlocked with one click.
[0086] Specifically, the security audit of the application security access control system includes system internal audit and user access behavior audit. Among them, the system internal audit is the audit log of corresponding management behaviors such as user management, application management, position management, and system management. In addition, the application security access control system uses video and logs to achieve a complete record of business system access behaviors. At the same time, based on the internal search engine, it provides a retrieval function for business operation behaviors, provides an efficient and accurate accountability mechanism for the subsequent security incidents, and provides a unified monitoring platform for administrators. User behavior audit includes: video audit, which records all user operations on the application system in video form. The video supports settings such as picture quality, frame interval, and compression ratio, which can greatly reduce the size of the video file generated by the graphic audit, and supports local storage and playback of the video playback; business module audit, when the user accesses the business system, the system will record the business module information involved in the user operation process, and at the same time, for the business module function interface, it can extract POST, GET, web page title, web page content, sensitive information and other functions; file audit, file upload / download behavior audit, including file name, business module, user, time, IP and other information; clipboard, copy, cut, and paste behavior audit of the front-end page content of the business system; abnormal application access, record and alarm the abnormal login, illegal login, unauthorized access and other behaviors of the user; and log search engine, search for related business module audit, file audit, clipboard audit content according to keywords, and associate the corresponding application session according to the search results. That is, by recording all the user's access behavior logs during the session in detail, providing data analysis capabilities for the audit day, and providing an audit analysis rule management interface, managers can customize analysis rules to complete the analysis and processing of audit data.
[0087] Here, the value of the application security access control system includes: unified management of internal user identities, that is, the application security access control system can achieve unified management of internal users without integration through master-slave account management, password policy management, master-slave account association, etc. If there is a willingness or integration, a standard interface can be provided to complete the docking integration; one-stop access to business systems to improve user experience, that is, the application security access control system provides a unified entrance for users to access business systems. Users no longer need to remember and enter the domain name / IP address port number of the business system and the matching username and password to access all business systems. They only need to log in to the portal system of the application fortress. After authentication, they can access on demand and sign in to the target system with a single sign-on. Business system; unified upgrade of authentication strength to meet the requirements of graded protection, that is, the application security access control system provides a unified authentication hub and a variety of strong authentication methods, including username and password, certificate, OTP, third-party interface docking authentication services, etc., supports multiple authentication methods, and can support different authentication methods for different users and different applications. The unified upgrade of business system authentication methods is achieved through the master-slave account association method, which reduces the cost of upgrading authentication construction while meeting the relevant requirements of graded protection on identity authentication; business system "invisibility" reduces business system security risks, that is, the application security access control system runs the managed business system in the business security container, and the application security access control system, All business systems are released to the outside in a unified way, and all access requests to the business system must and can only be authenticated, authorized, controlled and audited through the application fortress. All business systems do not expose any address or port information to the outside, which can effectively prevent illegal scanning and penetration from both inside and outside, and uniformly improve the security of all business systems; the minimum authorization that goes deep into the business logic, that is, the application security access control system can achieve the minimum functional level authorization less than the internal business without business transformation, and provide access control strategies based on security dimensions, such as setting access time, access location, access behavior and other information to deny access to specific sensitive functions. In special circumstances, you can apply for approval to temporarily increase permissions and access specific sensitive functions under controlled circumstances without affecting Ensure the data security of the business system without affecting the normal work of users; strengthen the protection capability of the business system, that is, the application security access control system does not need to transform the business system, that is, to achieve data desensitization, page watermarking, file upload and download / control of the business system, and through the terminal security container technology, control the clipboard and F12 debugging mode of the operating terminal, and conduct real-time monitoring and control of terminal operations to strengthen the data security protection capability of the business system. At the same time, through in-depth detection of application protocols, it can accurately match the attack characteristics, algorithms, and patterns of the application layer, and has certain application security detection and intrusion prevention capabilities, which can effectively resist attack scanning and hacker penetration. The page clear watermark technology can also effectively prevent phishing-like networks;Efficient encryption protocol conversion and business confidentiality protection, that is, for business systems that do not use SSL encryption, the server-side security container provides user access channel encryption capabilities without any modification to the business system, improving the security of business system access while meeting relevant policy requirements; unified audit of business access behavior and intelligent security risk analysis, that is, the system provides comprehensive security audit capabilities, filling the gap in the business system's audit of user business application operation behavior, which not only meets compliance requirements but also provides a data source for more advanced intelligent analysis. The system comprehensively records all user access behaviors to the business system, and can provide different levels of audit methods according to the differences in security audit requirements for different business scenarios, including terminal operation behavior video audit, and provides video positioning playback function to improve the availability of video audit. The system's intelligent security analysis engine can log in and authenticate users through user access Authentication, behavior operation and other logs, intelligent business chain association, context scenario analysis, so as to show the complete security map of the entire business system access, and display and alarm security events. In the security architecture of the security protection 2.0, the highest-level component is the security management center. Each security system needs to summarize the audit information to the security management center. The system makes up for the problem of insufficient audit logs in the business system. It can seamlessly connect to the security management center through multiple standard protocol interfaces, avoiding the problem of secondary development and adaptation of the business system; and encapsulate the client container and hide the commonly used ports, that is, encapsulate the client container, hide the commonly used ports 80, 8080, 8443, 443, etc., and use the customized unified encryption port 51000 for business communication, which can effectively prevent attacks and information theft through commonly used ports, greatly improving the security of the business system. ;
[0088] Figure 1 The following is an application scenario diagram of the application security access control system according to the embodiment of the present application. The application security access control system adopts physical bypass and logical series connection, does not change the user's existing network architecture, and is flexible and convenient to deploy. In addition, the system supports separate deployment and centralized deployment; supports dual machines and horizontal cluster expansion; highly productized delivery, convenient configuration, plug-and-play; provides plug-in-free mode, and is compatible with various terminals.
[0089] Exemplary Systems
[0090] Figure 2 FIG. 1 is a block diagram of an identity authentication and access control system based on security management according to an embodiment of the present application. Figure 2As shown, according to the embodiment of the present application, the identity authentication and access control system 100 based on security management includes: an access factor data monitoring unit 110, which is used to obtain the access factor data of the user to be detected, and the access factor data includes: user access time, location, sensitivity of the business system and access behavior; an access factor semantic understanding unit 120, which is used to pass the access factor data of the user to be detected through a semantic encoder including an embedding layer to obtain multiple context access factor semantic vectors; a cascade unit 130, which is used to cascade the multiple context access factor semantic vectors to obtain a global access semantic feature vector; a multi-scale semantic association feature extraction unit 140, which is used to arrange the multiple context access factor semantic vectors into feature vectors in one dimension and then pass them through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector; a multi-scale fusion unit 150, which is used to fuse the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain a classification feature vector; and an access control generation unit 160, which is used to pass the classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether to generate a secondary strong authentication.
[0091] Figure 3 FIG. 1 is an architecture diagram of an identity authentication and access control system based on security management according to an embodiment of the present application. Figure 3 As shown in the architecture diagram, firstly, the access factor data of the user to be detected is obtained, and the access factor data includes: user access time, location, sensitivity of the business system and access behavior. Then, the access factor data of the user to be detected is passed through a semantic encoder including an embedding layer to obtain a plurality of contextual access factor semantic vectors. Then, the plurality of contextual access factor semantic vectors are cascaded to obtain a global access semantic feature vector. Furthermore, the plurality of contextual access factor semantic vectors are arranged one-dimensionally into feature vectors and then passed through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector. Then, the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector are fused to obtain a classification feature vector. Then, the classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether a secondary strong authentication is generated.
[0092] In the above-mentioned identity authentication and access control system 100 based on security management, the access factor data monitoring unit 110 is used to obtain the access factor data of the user to be detected, and the access factor data includes: user access time, location, sensitivity of the business system and access behavior. As mentioned above, due to the relative independence of the construction time, construction department, construction manufacturer, etc. of multiple business systems of various enterprises and institutions, and the main construction requirements of most business systems focus more on the business logic processing perspective, there is a lack of unified management of multiple business systems and security access control considerations, and there are major management problems and security risks. Therefore, an optimized identity authentication and access control system based on security management is expected.
[0093] Accordingly, in the technical solution of this application, an application security access control system is constructed to provide comprehensive security protection and user access control for the business system through terminal security containers and server security containers. The system adopts a structural model in which the data plane and the control plane are separated from each other, ensuring that a large amount of data processing does not affect management and control. In a complex network environment, the control plane and the management plane do not affect each other, which highly guarantees the security and stability of the system. In addition, based on factors such as user access time, location, sensitivity of the business system, and access behavior, users are required to undergo secondary strong authentication to ensure that the visitor is a legal and authorized natural person.
[0094] However, during the secondary strong authentication process, the applicant of this application found that if the secondary strong authentication is inappropriate, it will greatly affect the user experience, and if the secondary strong authentication is not timely, it will reduce the security of access. In view of the above technical problems, it is expected to be able to intelligently determine whether to generate a secondary strong authentication based on the access factor data of the detected user.
[0095] Specifically, the access factor data of the detected user implies the access behavior characteristics of the detected user, but this mapping relationship is implicit and nonlinear, and it is difficult to accurately mine and understand based on traditional machine learning or feature engineering methods. Accordingly, in recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in the fields of image classification, object detection, semantic segmentation, text translation, etc. The development of deep learning and neural networks provides new solutions and solutions for understanding user access behavior.
[0096] Accordingly, in the technical solution of the present application, the access factor data of the user to be detected is first obtained, and the access factor data includes: user access time, location, sensitivity of the business system and access behavior.
[0097] In the above-mentioned identity authentication and access control system 100 based on security management, the access factor semantic understanding unit 120 is used to pass the access factor data of the user to be detected through a semantic encoder including an embedding layer to obtain multiple contextual access factor semantic vectors. Here, considering that each data in the access factor data is not structured data, before inputting the semantic understanding model for semantic understanding, each data item in the access factor data of the user to be detected is first structured. Specifically, the embedding layer of the context encoder is used to map each data item in the access factor data of the user to be detected to an embedding vector to obtain a sequence of embedded vectors. For example, it can be vector embedded through a learnable embedding matrix. Then, the context encoder is used to perform converter-based contextual semantic encoding on the sequence of embedded vectors to obtain the multiple contextual access factor semantic vectors.
[0098] More specifically, in an embodiment of the present application, the access factor semantic understanding unit 120 includes: an embedding unit, used to use the embedding layer of the context encoder to map each data item in the access factor data of the user to be detected into an embedding vector to obtain a sequence of embedding vectors; and a context encoding unit, used to use the context encoder to perform converter-based context semantic encoding on the sequence of embedding vectors to obtain the multiple context access factor semantic vectors.
[0099] More specifically, in an embodiment of the present application, the context encoding unit is further used to: arrange the sequence of the embedding vectors into an input vector; convert the input vector into a query vector and a key vector respectively through a learnable embedding matrix; calculate the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; normalize the self-attention association matrix to obtain a standardized self-attention association matrix; input the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and multiply the self-attention feature matrix with each embedding vector in the sequence of the embedding vectors as a value vector to obtain the multiple context access element semantic vectors.
[0100] In the above-mentioned identity authentication and access control system 100 based on security management, the cascading unit 130 is used to cascade the multiple context access element semantic vectors to obtain a global access semantic feature vector. Here, in order to fully explore the access behavior semantic information contained in the access element data, the multiple context access element semantic vectors obtained are further understood at a multi-scale. Specifically, the multiple context access element semantic vectors are first cascaded to obtain a global access semantic feature vector.
[0101] Specifically, in the embodiment of the present application, the cascading unit 130 is further used to: cascade the multiple context access element semantic vectors to obtain a global access semantic feature vector according to the following formula; wherein the formula is:
[0102] X=Concat[V1,V2,...,V k ]
[0103] Among them, V1, V2, ..., V k represents the semantic vector of each context access element, and Concat[·,·,·] represents the cascade function.
[0104] In the above-mentioned identity authentication and access control system 100 based on security management, the multi-scale semantic association feature extraction unit 140 is used to arrange the multiple context access element semantic vectors into feature vectors in one dimension and then pass them through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector. Similarly, in order to fully mine the access behavior semantic information contained in the access element data, the obtained multiple context access element semantic vectors are further understood at a multi-scale. Here, the multiple context access element semantic vectors are arranged into feature vectors in one dimension and then pass them through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector. Among them, the multi-scale neighborhood feature extraction module includes multiple parallel one-dimensional convolution layers, which can use one-dimensional convolution kernels with different scales to perform multi-scale one-dimensional convolution encoding on the feature vectors arranged by the multiple context access element semantic vectors to extract the local neighborhood semantic association features between different access elements in the multiple context access element semantic vectors to obtain the multi-scale neighborhood access semantic feature vector.
[0105] Specifically, in an embodiment of the present application, the multi-scale semantic association feature extraction unit 140 includes: a first-scale semantic association feature extraction subunit, used to input the feature vector into the first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first-scale neighborhood access semantic feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel of a first length; a second-scale semantic association feature extraction subunit, used to input the feature vector into the second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second-scale neighborhood access semantic feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and a multi-scale cascade subunit, used to cascade the first-scale neighborhood access semantic feature vector and the second-scale neighborhood access semantic feature vector to obtain the multi-scale neighborhood access semantic feature vector.
[0106] More specifically, in an embodiment of the present application, the first-scale semantic association feature extraction subunit is further used to: use the first convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the feature vector using the following formula to obtain the first-scale neighborhood access semantic feature vector; wherein the formula is:
[0107]
[0108] Wherein, a is the width of the first convolution kernel in the x direction, F(a) is the first convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the feature vector; the second-scale semantic association feature extraction subunit is further used to: use the second convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the feature vector according to the following formula to obtain the second-scale neighborhood access semantic feature vector; wherein the formula is:
[0109]
[0110] Among them, b is the width of the second convolution kernel in the x direction, F(b) is the second convolution kernel parameter vector, G(xb) is the local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, and X represents the eigenvector.
[0111] In the above-mentioned identity authentication and access control system 100 based on security management, the multi-scale fusion unit 150 is used to fuse the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain a classification feature vector. That is, after obtaining the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector, the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector are fused to obtain a classification feature vector. In this way, the access behavior semantic information contained in the access factor data is fully mined.
[0112] Specifically, in an embodiment of the present application, the multiscale fusion unit 150 is further used to calculate the position-weighted sum between the global access semantic feature vector and the multiscale neighborhood access semantic feature vector to obtain the classification feature vector. Here, in the position-weighted sum operation, the weight value is used to control the balance between the global access semantic feature vector and the multiscale neighborhood access semantic feature vector in the classification feature vector.
[0113] In the above-mentioned identity authentication and access control system 100 based on security management, the access control generation unit 160 is used to pass the classification feature vector through the classifier to obtain a classification result, and the classification result is used to indicate whether to generate a secondary strong authentication. In this way, based on the deep network model, the access behavior semantics of the access element data of the user to be detected are understood, and based on the extracted semantic features, it is determined whether to generate a secondary strong authentication, so as to improve the rationality and necessity of the generation of the secondary strong authentication and enhance the user experience.
[0114] Specifically, in the embodiment of the present application, the access control generation unit 160 is further used to: use the classifier to process the classification feature vector according to the following formula to obtain the classification result, wherein the formula is:
[0115] softmax{(W n ,B n ):…:(W1,B1)|X}
[0116] Among them, W1 to W n is the weight matrix, B1 to B n is the bias vector, and X is the classification feature vector.
[0117] The above-mentioned identity authentication and access control system 100 based on security management further includes a training module 200 for training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier.
[0118] Figure 4 FIG. 1 is a block diagram of a training module in an identity authentication and access control system based on security management according to an embodiment of the present application. Figure 4As shown, the training module 200 includes: a training data acquisition unit 210, which is used to acquire training data, wherein the training data includes access factor data of the training user to be detected, and a true value of whether the training user to be detected generates a secondary strong authentication; a training access factor semantic understanding unit 220, which is used to pass the access factor data of the training user to be detected through the semantic encoder including the embedding layer to obtain a plurality of training context access factor semantic vectors; a training cascade unit 230, which is used to cascade the plurality of training context access factor semantic vectors to obtain a training global access semantic feature vector; a training multi-scale semantic association feature extraction unit 240, which is used to arrange the plurality of training context access factor semantic vectors in one dimension into training feature vectors and then pass them through the multi-scale neighborhood feature extraction module to obtain to the training multi-scale neighborhood access semantic feature vector; a training multi-scale fusion unit 250, used to fuse the training global access semantic feature vector and the training multi-scale neighborhood access semantic feature vector to obtain a training classification feature vector; a classification loss unit 260, used to pass the training classification feature vector through the classifier to obtain a classification loss function value; an internalized learning loss unit 270, used to calculate the sequence-to-sequence response rule internalized learning loss function value between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector; and a training unit 280, used to train the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier with the weighted sum of the internalized learning loss function value and the classification loss function value as the loss function value pair.
[0119] Here, after the multiple context access element semantic vectors are arranged in one dimension as feature vectors and then a multi-scale neighborhood access semantic feature vector is obtained through a multi-scale neighborhood feature extraction module, it is expected that the multi-scale neighborhood feature extraction module can extract the associated features of the context access element semantics under the multi-scale neighborhood while still having good responsiveness to the global access semantic feature vector obtained by directly cascading the multiple context access element semantic vectors, so as to enhance the fusion effect of the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector. Therefore, the sequence-to-sequence response rule internalization learning loss function between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector is calculated.
[0120] Specifically, in an embodiment of the present application, the internalized learning loss unit is further used to calculate the sequence-to-sequence response rule internalized learning loss function between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector using the following formula; wherein the formula is:
[0121]
[0122]
[0123]
[0124] Among them, V1 and V2 are the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector respectively, and W1 and W2 are the weight matrices of the classifier for V1 and V2 respectively, and d represents the Euclidean distance between the two vectors.
[0125] That is, the enhanced distinguishing ability between sequences is obtained by using the squeeze-excite channel attention mechanism of the classifier for the weight matrices of different vector sequences represented by the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector. In this way, by training the network with this loss function, the recovery of causal relationship features with better distinguishing characteristics between response sequences can be achieved, so as to internalize the cause-effect response rules between sequences and improve the responsiveness of the multi-scale neighborhood access semantic feature vector to the global access semantic feature vector.
[0126] In summary, according to the security management-based identity authentication and access control system 100 of the embodiment of the present application, it is explained that it takes into account that the access factor data of the detected user implies the access behavior characteristics of the detected user. Therefore, a semantic encoder including an embedding layer is used to mine the global implicit associations between the various access factor data to obtain multiple contextual access factor semantic vectors, and further, a multi-scale understanding is performed on the obtained multiple contextual access factor semantic vectors. In this way, the access behavior semantics of the access factor data of the user to be detected is understood based on the deep network model, and whether to generate a secondary strong authentication is determined based on the extracted semantic features, so as to improve the rationality and necessity of generating the secondary strong authentication.
[0127] Exemplary Methods
[0128] Figure 5 FIG. 1 is a flow chart of an identity authentication and access control method based on security management according to an embodiment of the present application. Figure 5As shown, according to the security management-based identity authentication and access control method of the embodiment of the present application, the steps include: S110, obtaining access factor data of the user to be detected, the access factor data including: user access time, location, sensitivity of the business system and access behavior; S120, passing the access factor data of the user to be detected through a semantic encoder including an embedding layer to obtain multiple context access factor semantic vectors; S130, cascading the multiple context access factor semantic vectors to obtain a global access semantic feature vector; S140, arranging the multiple context access factor semantic vectors into feature vectors in one dimension and then passing them through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector; S150, fusing the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain a classification feature vector; and, S160, passing the classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether to generate a secondary strong authentication.
[0129] In one example, in the above-mentioned identity authentication and access control method based on security management, the access factor data of the user to be detected is passed through a semantic encoder including an embedding layer to obtain multiple contextual access factor semantic vectors, including: using the embedding layer of the context encoder to map each data item in the access factor data of the user to be detected into an embedding vector to obtain a sequence of embedded vectors; and, using the context encoder to perform converter-based contextual semantic encoding on the sequence of embedded vectors to obtain the multiple contextual access factor semantic vectors.
[0130] In one example, in the above-mentioned security management-based identity authentication and access control method, the use of the context encoder to perform converter-based context semantic encoding on the sequence of embedding vectors to obtain the multiple context access element semantic vectors includes: arranging the sequence of embedding vectors as input vectors; converting the input vectors into query vectors and key vectors respectively through learnable embedding matrices; calculating the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; normalizing the self-attention association matrix to obtain a standardized self-attention association matrix; inputting the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and multiplying the self-attention feature matrix with each embedding vector in the sequence of embedding vectors as a value vector to obtain the multiple context access element semantic vectors.
[0131] In one example, in the above-mentioned identity authentication and access control method based on security management, it also includes training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier.
[0132] Figure 6 The flowchart of training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier in the identity authentication and access control method based on security management according to the embodiment of the present application is as follows. Figure 6 As shown, the process of training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier includes the steps of: S210, obtaining training data, the training data including access factor data of the training user to be detected, and a true value of whether the training user to be detected generates a secondary strong authentication; S220, passing the access factor data of the training user to be detected through the semantic encoder including the embedding layer to obtain a plurality of training context access factor semantic vectors; S230, cascading the plurality of training context access factor semantic vectors to obtain a training global access semantic feature vector; S240, arranging the plurality of training context access factor semantic vectors in one dimension into training feature vectors and then passing the plurality of training context access factor semantic vectors through the multi-scale neighborhood feature extraction module. A scale neighborhood feature extraction module is used to obtain a training multi-scale neighborhood access semantic feature vector; S250, the training global access semantic feature vector and the training multi-scale neighborhood access semantic feature vector are integrated to obtain a training classification feature vector; S260, the training classification feature vector is passed through the classifier to obtain a classification loss function value; S270, the sequence-to-sequence response rule internalized learning loss function value between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector is calculated; and, S280, the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier are trained with the weighted sum of the internalized learning loss function value and the classification loss function value as the loss function value pair.
[0133] In summary, the identity authentication and access control method based on security management according to the embodiment of the present application is explained, which takes into account that the access factor data of the detected user implies the access behavior characteristics of the detected user. Therefore, a semantic encoder including an embedding layer is used to mine the global implicit associations between the various access factor data to obtain multiple contextual access factor semantic vectors, and further, a multi-scale understanding is performed on the obtained multiple contextual access factor semantic vectors. In this way, the access behavior semantics of the access factor data of the user to be detected is understood based on the deep network model, and whether to generate a secondary strong authentication is determined based on the extracted semantic features, so as to improve the rationality and necessity of generating the secondary strong authentication.
Claims
1. An identity authentication and access control system based on security management, characterized in that: include: An access factor data monitoring unit is used to obtain access factor data of the user to be detected, wherein the access factor data includes: user access time, location, sensitivity of the business system and access behavior; An access factor semantic understanding unit, used for passing the access factor data of the user to be detected through a semantic encoder including an embedding layer to obtain a plurality of contextual access factor semantic vectors; A cascading unit, configured to cascade the plurality of context access element semantic vectors to obtain a global access semantic feature vector; A multi-scale semantic association feature extraction unit, used for one-dimensionally arranging the plurality of context access element semantic vectors into feature vectors and then passing them through a multi-scale neighborhood feature extraction module to obtain a multi-scale neighborhood access semantic feature vector; a multi-scale fusion unit, configured to fuse the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain a classification feature vector; and An access control generation unit, used for passing the classification feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether to generate a secondary strong authentication; Wherein, the access element semantic understanding unit includes: an embedding unit, configured to use the embedding layer of the context encoder to map each data item in the access factor data of the to-be-detected user into an embedding vector to obtain a sequence of embedding vectors; and A context encoding unit, configured to perform transformer-based context semantic encoding on the sequence of the embedding vectors using the context encoder to obtain the plurality of context access element semantic vectors; Wherein, the context encoding unit is further used for: Arranging the sequence of embedding vectors as an input vector; Convert the input vector into a query vector and a key vector respectively through a learnable embedding matrix; Calculating the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; Normalizing the self-attention association matrix to obtain a standardized self-attention association matrix; Inputting the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and Multiplying the self-attention feature matrix by each embedding vector in the sequence of embedding vectors as a value vector to obtain the plurality of context access element semantic vectors; Wherein, the cascade unit is further used for: The multiple context access element semantic vectors are concatenated according to the following formula to obtain a global access semantic feature vector; Wherein, the formula is: X=Concat[V1,V2,...,V k ] Among them, V1, V2, ..., V k represents the semantic vector of each context access element, and Concat[·,·,·] represents the cascade function.
2. The identity authentication and access control system based on security management according to claim 1, characterized in that: The multi-scale semantic association feature extraction unit comprises: A first-scale semantic association feature extraction subunit, configured to input the feature vector into a first convolutional layer of the multi-scale neighborhood feature extraction module to obtain a first-scale neighborhood access semantic feature vector, wherein the first convolutional layer has a first one-dimensional convolutional kernel of a first length; a second-scale semantic association feature extraction subunit, configured to input the feature vector into a second convolutional layer of the multi-scale neighborhood feature extraction module to obtain a second-scale neighborhood access semantic feature vector, wherein the second convolutional layer has a second one-dimensional convolutional kernel of a second length, and the first length is different from the second length; and The multi-scale cascading subunit is used to cascade the first-scale neighborhood access semantic feature vector and the second-scale neighborhood access semantic feature vector to obtain the multi-scale neighborhood access semantic feature vector.
3. The identity authentication and access control system based on security management according to claim 2, characterized in that: The first-scale semantic association feature extraction subunit is further used to: use the first convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the feature vector according to the following formula to obtain the first-scale neighborhood access semantic feature vector; Wherein, the formula is: Wherein, a is the width of the first convolution kernel in the x direction, F(a) is the first convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the eigenvector; The second-scale semantic association feature extraction subunit is further used to: use the second convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the feature vector according to the following formula to obtain the second-scale neighborhood access semantic feature vector; Wherein, the formula is: Among them, b is the width of the second convolution kernel in the x direction, F(b) is the second convolution kernel parameter vector, G(xb) is the local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, and X represents the eigenvector.
4. The identity authentication and access control system based on security management according to claim 3, characterized in that: The multi-scale fusion unit is further used to calculate the position-weighted sum between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector to obtain the classification feature vector.
5. The identity authentication and access control system based on security management according to claim 4, characterized in that: The access control generation unit is further used to: The classification feature vector is processed by the classifier using the following formula to obtain the classification result, wherein the formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is the bias vector, and X is the classification feature vector.
6. The identity authentication and access control system based on security management according to claim 1, characterized in that: Also included is a training module for training the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module, and the classifier; Wherein, the training module includes: A training data acquisition unit, used to acquire training data, wherein the training data includes access factor data of the training user to be detected and a true value of whether a secondary strong authentication is generated for the training user to be detected; A training access element semantic understanding unit, used for passing the access element data of the training to-be-detected user through the semantic encoder including the embedding layer to obtain a plurality of training context access element semantic vectors; A training cascade unit, configured to cascade the plurality of training context access element semantic vectors to obtain a training global access semantic feature vector; A training multi-scale semantic association feature extraction unit, used for one-dimensionally arranging the plurality of training context access element semantic vectors into a training feature vector and then passing the vector through the multi-scale neighborhood feature extraction module to obtain a training multi-scale neighborhood access semantic feature vector; A training multi-scale fusion unit, used for fusing the training global access semantic feature vector and the training multi-scale neighborhood access semantic feature vector to obtain a training classification feature vector; A classification loss unit, used for passing the training classification feature vector through the classifier to obtain a classification loss function value; an internalized learning loss unit, used to calculate a sequence-to-sequence response rule internalized learning loss function value between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector; and A training unit is used to train the semantic encoder including the embedding layer, the multi-scale neighborhood feature extraction module and the classifier by using the weighted sum of the internalized learning loss function value and the classification loss function value as the loss function value pair.
7. The identity authentication and access control system based on security management according to claim 6, characterized in that: The internalized learning loss unit is further used to: The sequence-to-sequence response rule internalization learning loss function between the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector is calculated by the following formula; Wherein, the formula is: Among them, V1 and V2 are the global access semantic feature vector and the multi-scale neighborhood access semantic feature vector respectively, and W1 and W2 are the weight matrices of the classifier for V1 and V2 respectively.
Citation Information
Patent Citations
User browsing behavior authentication method and system
CN107368718A
Equipment real-time performance evaluation system and method based on multi-dimensional data collaboration
CN115564203A
