Attack detection model optimization and attack detection method, device and storage medium

Abstract

The embodiment of the application discloses an attack detection model optimization and attack detection method, device and storage medium, relates to the technical field of network security, and comprises the following steps: in response to satisfying a preset training trigger condition, determining the conditional constraint of a misjudgment sample of a previous attack generation model; using the previous attack generation model, generating attack sample features of corresponding attack type labels based on the previous potential attack vector and the conditional constraint of each misjudgment sample, and using an attacker to generate an attack sample corresponding to each attack sample feature; updating the previous voting weight combination and the previous discrimination loss function based on a plurality of misjudgment samples to obtain a current voting weight combination and a current discrimination loss function; and training and optimizing the previous attack detection model based on the current voting weight combination, the current discrimination loss function, the plurality of attack samples and the attack type label corresponding to each attack sample to obtain a current attack detection model, so that the attack detection accuracy in a complex attack scene can be improved.

Description

Technical Field

[0001] This application relates to the field of network security technology, and in particular to an attack detection model optimization and attack detection method, device and storage medium. Background Technology

[0002] Currently, detection methods based on static rules or feature matching typically identify the attack type in the actual operating data of a computer to determine whether the computer has been maliciously attacked. However, for complex attack scenarios where attack characteristics are modified through means such as deformation, encryption, and obfuscation, the above solutions cannot accurately identify the attack type, resulting in low attack detection accuracy in complex attack scenarios. Summary of the Invention

[0003] This application provides an optimized attack detection model and an attack detection method, device, and storage medium, which can improve the accuracy of attack detection in complex attack scenarios, thereby solving the problem of low attack detection accuracy in complex attack scenarios in the prior art.

[0004] In a first aspect, embodiments of this application provide an attack detection model optimization method, which includes: in response to satisfying a preset training trigger condition, acquiring multiple misjudged samples of the previous attack detection model and the attack type label corresponding to each misjudged sample, and determining the key conditions of the previous attack detection model when identifying each misjudged sample, thereby obtaining the condition constraints of the corresponding misjudged sample; using the previous attack generation model, based on the previous potential attack vector and the condition constraints of each misjudged sample, generating attack sample features corresponding to the attack type label, and using an attacker to generate attack samples corresponding to the attack sample features based on each attack sample feature; updating the previous voting weight combination and the previous discriminant loss function of the previous attack detection model based on multiple misjudged samples, thereby obtaining the current voting weight combination and the current discriminant loss function, and training and optimizing the previous attack detection model based on the current voting weight combination, the current discriminant loss function, multiple attack samples, and the attack type label corresponding to each attack sample, thereby obtaining the current attack detection model.

[0005] Secondly, embodiments of this application provide an attack detection method, which includes: acquiring actual operating data at the current moment; inputting the actual operating data into a current attack detection model obtained by using the attack detection model optimization method of any embodiment of this application to obtain an attack detection result; and generating and displaying warning information based on the attack detection result when the attack detection result is within a preset attack type list.

[0006] Thirdly, embodiments of this application provide an attack detection model optimization apparatus, the apparatus comprising:

[0007] The determination module is used to obtain multiple misjudged samples of the previous attack detection model and the attack type label corresponding to each misjudged sample in response to the satisfaction of the preset training trigger conditions, and to determine the key conditions of the previous attack detection model when identifying each misjudged sample, so as to obtain the condition constraints of the corresponding misjudged sample.

[0008] The generation module is used to generate attack sample features with corresponding attack type labels based on the previous attack generation model, the previous potential attack vector and the conditional constraints of each misjudged sample, and the attacker generates attack samples with corresponding attack sample features based on each attack sample feature.

[0009] The first training module is used to update the previous voting weight combination and the previous discriminant loss function of the previous attack detection model based on multiple misjudged samples, thereby obtaining the current voting weight combination and the current discriminant loss function. Based on the current voting weight combination, the current discriminant loss function, multiple attack samples, and the attack type label corresponding to each attack sample, the previous attack detection model is trained and optimized to obtain the current attack detection model.

[0010] Fourthly, embodiments of this application provide an attack detection device, which includes:

[0011] The acquisition module is used to acquire the actual running data at the current moment;

[0012] An attack detection module is used to input actual running data into the current attack detection model obtained by the attack detection model optimization method of any embodiment of this application, and obtain the attack detection result.

[0013] The early warning module is used to generate and display early warning information based on the attack detection results when the attack detection results are within the preset attack type list.

[0014] Fifthly, embodiments of this application provide an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the attack detection model optimization method of any embodiment of this application, or to enable the at least one processor to execute the attack detection method of any embodiment of this application.

[0015] Sixthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon. When executed by a processor, the program implements the attack detection model optimization method as described in any embodiment of this application, or when executed by a processor, the program implements the attack detection method as described in any embodiment of this application.

[0016] In this embodiment, in response to the satisfaction of a preset training trigger condition, multiple misjudged samples from the previous attack detection model and the attack type label corresponding to each misjudged sample can be obtained. At this time, the misjudged samples can actively expose the blind spots of the previous attack detection model, providing an accurate data foundation for subsequent training of the previous attack detection model. Secondly, the key conditions of the previous attack detection model in recognizing each misjudged sample are determined, and the condition constraints of the corresponding misjudged samples are obtained. Then, using the previous attack generation model, based on the previous potential attack vector and the condition constraints of each misjudged sample, attack sample features corresponding to the attack type label are generated. Through the condition constraints of the misjudged samples, the previous attack generation model can generate attack sample features that bypass the condition constraints, avoiding the generation of invalid or out-of-range attack sample features. The attacker then uses the attack sample generator to generate attack samples with corresponding attack sample features based on each attack sample feature. At this time, the attack samples are difficult for the previous attack detection model to recognize. Then, based on multiple misjudged samples... The previous voting weight combination and discriminant loss function of the previous attack detection model are updated to obtain the current voting weight combination and current discriminant loss function. Based on the current voting weight combination, current discriminant loss function, multiple attack samples, and the attack type label corresponding to each attack sample, the previous attack detection model is trained and optimized to obtain the current attack detection model. The bypass techniques of the attack generation model (i.e., attackers) (such as techniques to modify attack features by deformation, encryption, obfuscation, etc.) can be transformed into evolution points for the attack detection model (i.e., defenders). This achieves dynamic enhancement and adaptive improvement of the attack detection model, enabling the current attack detection model to accurately identify the attack type corresponding to the attack sample adjusted by the bypass technique. This further improves the recognition accuracy of the current attack detection model, especially in complex attack scenarios where attack features are modified by deformation, encryption, obfuscation, etc., thereby improving the attack detection accuracy of the current attack detection model in complex attack scenarios. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the attack detection model optimization method provided in an embodiment of this application;

[0019] Figure 2 This is another flowchart illustrating the attack detection model optimization method provided in this application embodiment;

[0020] Figure 3This is a flowchart illustrating an attack detection method provided in an embodiment of this application;

[0021] Figure 4 This is a schematic diagram of the attack detection model optimization device provided in an embodiment of this application;

[0022] Figure 5 This is a schematic diagram of the attack detection device provided in an embodiment of this application;

[0023] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0025] It should be noted that the terms "first," "second," "target," and "original," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "comprising," "having," and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] Figure 1 This is a flowchart illustrating an attack detection model optimization method provided in this application. This embodiment can be applied to scenarios where it is necessary to improve the accuracy of attack detection models in complex attack scenarios. The attack detection model optimization method provided in this embodiment can be executed by the attack detection model optimization device provided in this application, which can be implemented through software and / or hardware. In a specific embodiment, the attack detection model optimization device can be integrated into an electronic device, such as a computer. The executing entity of this method can be an electronic device. See also... Figure 1 The attack detection model optimization method in this embodiment includes, but is not limited to, the following steps:

[0027] S110. In response to the satisfaction of the preset training triggering conditions, obtain multiple misjudged samples of the previous attack detection model and the attack type label corresponding to each misjudged sample, and determine the key conditions of the previous attack detection model when identifying each misjudged sample, and obtain the condition constraints of the corresponding misjudged sample.

[0028] The preset training trigger condition is a pre-set training trigger condition for the attack detection model. For example, the preset training trigger condition may include a time difference between the current time and the previous training time being greater than a preset duration; the previous training time is the time when the attack detection model was last trained; the preset duration is a pre-set duration used to characterize the training cycle of the attack detection model.

[0029] The attack detection model is a classification model used to identify the attack type corresponding to the attack sample. For example, the attack detection model can be a random forest model. The attack sample is a specific instance in the field of network security that can reflect the characteristics of malicious attack behavior, i.e., attack traffic. It can be an executable file containing malicious code or a network data packet used to launch an attack. The attack type is the network attack type corresponding to the attack sample. The attack type label is the correct attack type corresponding to the misjudged sample.

[0030] The previous attack detection model is the attack detection model obtained from the most recent training. False positives are attack samples that the previous attack detection model incorrectly classified. Optionally, if the attack detection model's predicted probability of classifying an attack sample as a certain type is less than a preset probability threshold, the attack sample is considered a false positive. The preset probability threshold is a pre-set probability value used to determine false positives.

[0031] The key conditions are those that have a decisive or significant impact on the final classification result during the classification and identification process of the previous attack detection model; the condition constraints are the key conditions.

[0032] Specifically, when the time difference between the current moment and the previous training moment is detected to be greater than a preset duration, it is determined that the preset training trigger condition is met. At this time, multiple misjudged samples from the previous attack detection model and the attack type label corresponding to each misjudged sample can be obtained. Here, the misjudged samples are the attack samples corresponding to the attack sample features generated by the previous attack generation model. The key conditions of the previous attack detection model in identifying each misjudged sample are determined, and the condition constraints of the corresponding misjudged sample are obtained. For example, for the current misjudged sample among multiple misjudged samples, the discrimination condition path of the previous attack detection model in identifying the current misjudged sample can be determined, and multiple test paths are set based on all the nodes included in the discrimination condition path. Then, the multiple test paths are tested to determine which node in the discrimination condition path is the key node. Then, the conditions corresponding to the key node are determined as key conditions, and the key conditions are determined as the condition constraints corresponding to the current misjudged sample.

[0033] The attack generation model is a generative neural network model used to generate attack sample features; the previous attack generation model is the attack generation model obtained from the most recent training; the attack sample features are the key attack features corresponding to the attack sample, i.e., the attack payload; the discrimination condition path is a logical judgment route formed by the nodes that the previous attack detection model passes through in sequence when classifying the attack sample, and each node corresponds to one condition.

[0034] S120. Using the previous attack generation model, based on the previous potential attack vector and the conditional constraints of each misjudged sample, generate attack sample features corresponding to the attack type label, and use the attacker to generate attack samples corresponding to the attack sample features based on each attack sample feature.

[0035] The potential attack vector represents the attacker's intended attack and related constraints, such as attack strength and perturbation range. Each dimension of the potential attack vector corresponds to different attack-related attributes or features. The previous potential attack vector is the one obtained in the most recent training.

[0036] The attacker is an engine used to combine the features of attack samples into attack samples. It is used to generate attack samples corresponding to the features of attack samples based on real attack rules, thereby simulating real network attack behavior.

[0037] Specifically, the previous potential attack vector and the previous attack generation model can be obtained. Then, for the current misjudged sample among multiple misjudged samples, the previous potential attack vector and the conditional constraints corresponding to the current misjudged sample can be input into the previous attack generation model. At this time, the previous attack generation model can generate attack sample features that bypass the conditional constraints corresponding to the current misjudged sample. That is, the attack sample features can maintain the destructiveness of the original attack, but avoid the detection of the conditional constraints corresponding to the current misjudged sample. In other words, the attack sample features are difficult to be identified by the conditional constraints corresponding to the current misjudged sample, thus making it difficult for the previous attack detection model to identify the attack sample features. Moreover, the attack type label corresponding to this attack sample feature is the attack type label corresponding to the current misjudged sample. Then, the attack tool is used to generate the attack sample corresponding to the attack sample feature, that is, the attack traffic. Here, the attack sample is the attack sample that the previous attack detection model is difficult to identify, providing an accurate data foundation for the subsequent training of the previous attack detection model. And the attack type label corresponding to this attack sample is the attack type label corresponding to the current misjudged sample.

[0038] S130. Based on multiple misjudged samples, update the previous voting weight combination and the previous discriminant loss function of the previous attack detection model to obtain the current voting weight combination and the current discriminant loss function. Then, based on the current voting weight combination, the current discriminant loss function, multiple attack samples, and the attack type label corresponding to each attack sample, train and optimize the previous attack detection model to obtain the current attack detection model.

[0039] The voting weight combination includes the voting weights for each attack type, reflecting the influence of each attack type on the final classification result. The previous voting weight combination is the voting weight combination obtained in the most recent update. The current voting weight combination is the voting weight combination obtained after updating the previous voting weight combination based on multiple misclassified samples.

[0040] The discriminant loss function is the loss function of the attack detection model; the previous discriminant loss function is the discriminant loss function obtained from the most recent update. The current discriminant loss function is the discriminant loss function obtained after updating the previous discriminant loss function based on multiple misclassified samples. The current attack detection model is the attack detection model obtained after training and optimizing the previous attack detection model.

[0041] Specifically, the previous voting weight combination and the previous discriminant loss function of the previous attack detection model can be updated based on multiple misjudged samples. For example, the recall rate of each attack type can be determined based on multiple misjudged samples, and then the previous voting weight combination can be updated based on the recall rate of each attack type to obtain the current voting weight combination. The previous discriminant loss function can also be updated based on the recall rate of each attack type to obtain the current discriminant loss function.

[0042] Then, for the current attack sample among multiple attack samples, the attacker simulates computer attack operations based on the current attack sample to launch an attack on the electronic device, and collects the operating data of the electronic device in real time, thereby obtaining the operating data corresponding to each attack sample among multiple attack samples. Next, each operating data in the multiple operating data is input into the previous attack detection model. At this time, the previous attack detection model uses the current voting weight combination to determine the judgment type of the corresponding attack sample. Then, the current discriminant loss function is used to calculate the discriminant loss value of the corresponding attack sample based on the attack type label and judgment type of each attack sample. At this time, the smaller the difference between the judgment type and the attack type label, the smaller the corresponding discriminant loss value. Then, with the goal of minimizing the discriminant loss value, the previous attack detection model is trained and optimized to obtain the current attack detection model, so that the current attack detection model can accurately identify the attack type corresponding to the attack sample generated by the attack generation model, thereby continuously strengthening the recognition ability of the attack detection model and improving the recognition accuracy of the attack detection model.

[0043] Optionally, after obtaining the current attack detection model, the model parameters of the current attack detection model can be fixed, and the previous attack generation model can be trained and optimized. That is, each attack sample can be identified based on the current attack detection model to obtain the corresponding attack sample's judgment type. Specifically, the running data corresponding to each attack sample can be input into the current attack detection model, at which point the current attack detection model uses the current voting weight combination to determine the judgment type of the corresponding attack sample. Secondly, the previous attack generation model can be trained and optimized based on the judgment type and attack type label of each attack sample to obtain the current attack generation model. Specifically, the generation loss function of the previous attack generation model can be obtained, where the generation loss function is the attack type label. The loss function of the attack generation model is then used to calculate the generation loss value of each attack sample based on its judgment type and attack type label. The greater the difference between the judgment type and the attack type label, the smaller the corresponding generation loss value. The previous attack generation model is then trained and optimized to minimize this generation loss value. This process updates the previous potential attack vector, resulting in the current attack generation model. This allows the current attack generation model to generate attack sample features that effectively bypass the attack detection model (i.e., more adversarial attack sample features). In other words, it enables the attack generation model to generate attack sample features that are increasingly difficult to identify, thereby improving its attack sample generation capability. Through multiple rounds of alternating training and optimization between the attack detection model and the attack generation model, the attack generation model can generate attack sample features that are increasingly difficult for the attack detection model to identify. This forces the attack detection model to gradually strengthen its ability to identify these attack sample features, thus gradually enhancing both its type identification capability and its attack sample generation capability.

[0044] The technical solution of this application embodiment, in response to the satisfaction of preset training trigger conditions, can obtain multiple misjudged samples of the previous attack detection model and the attack type label corresponding to each misjudged sample. At this time, the misjudged samples can actively expose the recognition blind spots of the previous attack detection model, providing an accurate data foundation for subsequent training of the previous attack detection model; secondly, determine the key conditions of the previous attack detection model when recognizing each misjudged sample, and obtain the condition constraints of the corresponding misjudged sample. Then, using the previous attack generation model, based on the previous potential attack vector and the condition constraints of each misjudged sample, generate attack sample features corresponding to the attack type label. Through the condition constraints of the misjudged samples, the previous attack generation model can generate attack sample features that bypass the condition constraints, avoiding the generation of invalid or out-of-range attack sample features, and use the attacker to generate attack samples with corresponding attack sample features based on each attack sample feature. At this time, the attack samples are difficult to be recognized by the previous attack detection model; then, based on multiple misjudged samples... The sample updates the previous voting weight combination and the previous discriminant loss function of the previous attack detection model, resulting in the current voting weight combination and the current discriminant loss function. Based on the current voting weight combination, the current discriminant loss function, multiple attack samples, and the attack type label corresponding to each attack sample, the previous attack detection model is trained and optimized to obtain the current attack detection model. The bypass techniques of the attack generation model (i.e., attackers) (such as techniques that modify attack features through deformation, encryption, obfuscation, etc.) can be transformed into evolution points for the attack detection model (i.e., defenders), thereby realizing the dynamic enhancement and adaptive improvement of the attack detection model. This enables the current attack detection model to accurately identify the attack type corresponding to the attack sample adjusted by the bypass technique, further improving the recognition accuracy of the current attack detection model, especially in complex attack scenarios where attack features are modified through deformation, encryption, obfuscation, etc., thereby improving the attack detection accuracy of the current attack detection model in complex attack scenarios.

[0045] The following further describes an attack detection model optimization method provided by an embodiment of this application. Figure 2 This is another flowchart illustrating the attack detection model optimization method provided in this application. This application's embodiments are optimizations based on the above embodiments. See also... Figure 2 The method in this embodiment includes, but is not limited to, the following steps:

[0046] S210. In response to the satisfaction of the preset training triggering conditions, obtain multiple misjudged samples of the previous attack detection model and the attack type label corresponding to each misjudged sample, and determine the key conditions of the previous attack detection model when identifying each misjudged sample, and obtain the condition constraints of the corresponding misjudged sample.

[0047] Optionally, the preset training trigger condition may include the number of newly added misjudged samples being greater than a preset number threshold; wherein, the newly added misjudged samples are the misjudged samples of the previous attack detection model; the preset number threshold is a pre-set value used to represent the minimum value at which the number of newly added misjudged samples meets the preset training trigger condition.

[0048] Specifically, when the number of newly added misclassified samples exceeds a preset threshold, a preset training trigger condition is determined. At this point, multiple misclassified samples from the previous attack detection model and the attack type label corresponding to each misclassified sample can be obtained. Then, for the current misclassified sample among the multiple misclassified samples, the discrimination condition path of the previous attack detection model when identifying the current misclassified sample can be determined, and the condition corresponding to the node with the minimum information entropy in the discrimination condition path is determined as the key condition of the current misclassified sample. The node with the minimum information entropy is the key node in the discrimination condition path, thus obtaining the condition constraint corresponding to the current misclassified sample. This can improve computational efficiency, reduce implementation complexity, and thus improve the accuracy and efficiency of determining the key conditions, providing an accurate data foundation for subsequently generating attack sample features with corresponding attack type labels.

[0049] S220. Using the previous attack generation model, based on the previous potential attack vector and the conditional constraints of each misjudged sample, generate attack sample features corresponding to the attack type label, and use the attacker to generate attack samples corresponding to the attack sample features based on each attack sample feature.

[0050] S230. Determine the recall rate for each attack type based on multiple false positive samples.

[0051] Specifically, for the current attack type in each attack type, the number of misjudged samples with the attack type label of the current attack type can be determined among multiple misjudged samples, thus obtaining the current misjudged sample count corresponding to the current attack type. The total number of attack samples of the current attack type required to accumulate the current misjudged sample count of the current attack type can be obtained, that is, the total number of attack samples required to accumulate the current misjudged sample count can be obtained, thus obtaining the current total number of attack samples. Then, the difference between the current total number of attack samples and the current misjudged sample count is calculated to obtain the current correct sample count, and the ratio of the current correct sample count to the current total number of attack samples is calculated to obtain the recall rate of the current attack type.

[0052] S240. Update the previous voting weight combination based on the recall rate of each attack type to obtain the current voting weight combination, and update the previous discriminant loss function based on the recall rate of each attack type to obtain the current discriminant loss function.

[0053] Specifically, the previous voting weight combination is updated based on the recall rate of each attack type to obtain the current voting weight combination, including Sa1-Sa3:

[0054] Sa1: Obtain the preset base weights for each attack type.

[0055] Among them, the preset base weight is the initial voting weight set in advance for the attack type.

[0056] Sa2. Based on the previous discrimination learning rate of the previous attack detection model, the recall rate of each attack type, and the preset basic weights, determine the current voting weight of the corresponding attack type.

[0057] Here, the discriminant learning rate is the learning rate of the attack detection model; the previous discriminant learning rate is the discriminant learning rate obtained from the most recent training. The current voting weight is a voting weight calculated based on the previous discriminant learning rate, recall rate, and preset base weights.

[0058] Specifically, for the current attack type among all attack types, the difference between 1 and the recall rate of the current attack type can be calculated, and the product of this difference and the previous discrimination learning rate can be calculated. Then, the sum of this product and the preset basic weight of the current attack type is calculated to obtain the current voting weight of the current attack type. This can increase the current voting weight of attack types with low recall, thereby improving the sensitivity and accuracy of the attack detection model in identifying attack types with low recall.

[0059] Sa3. Update the previous voting weight combination based on the current voting weight of each attack type to obtain the current voting weight combination.

[0060] Specifically, the current voting weights of each attack type can be combined to obtain the current voting weight combination, and the previous voting weight combination can be updated based on the current voting weights of each attack type.

[0061] In this embodiment, the voting weight of attack types with low recall can be increased, and the computational efficiency can be improved while the implementation complexity can be reduced. This improves the accuracy and efficiency of determining the current voting weight combination, providing an accurate data foundation for subsequent training of the previous attack detection model.

[0062] Specifically, the previous discrimination loss function is updated based on the recall rate of each attack type to obtain the current discrimination loss function, including Sb1-Sb2:

[0063] Sb1. Based on the previous balance coefficient of the previous attack detection model and the recall rate of each attack type, determine the current category of sensitive data for the corresponding attack type.

[0064] The balance coefficient is used to quantify the attention of different attack types in the loss function, and different attack types correspond to the same balance coefficient; the previous balance coefficient is the balance coefficient obtained in the most recent training.

[0065] Category-sensitive data is used to give differentiated attention to different attack types, enabling the attack detection model to improve its learning of attack types with high category sensitivity during training and optimization; the current category-sensitive data is the category-sensitive data calculated based on the previous balance coefficient and recall rate.

[0066] Specifically, for the current attack type in each attack type, the difference between 1 and the recall rate of the current attack type can be calculated, and the square of the difference can be multiplied by the previous balance coefficient to obtain the current category sensitive data of the current attack type. This can improve the current category sensitive data of attack types with low recall, thereby improving the learning degree of the attack detection model for attack types with low recall, and thus improving the accuracy of the attack detection model in identifying attack types with low recall.

[0067] Sb2. Based on the current category sensitive data of each attack type, update the previous category sensitive data of the corresponding attack type in the previous discriminative loss function to obtain the current discriminative loss function.

[0068] Among them, the sensitive data of the previous category are the sensitive data of the categories included in the previous discriminative loss function.

[0069] Specifically, the current category sensitive data for each attack type can be replaced with the previous category sensitive data for the corresponding attack type in the previous discriminative loss function to obtain the current discriminative loss function.

[0070] Optionally, the formula for the discriminant loss function is: Where i represents attack type i, i∈{1,2,…,C}, and C represents the number of attack types; 1 i＝y It is an indicator function, where y represents the attack type label of attack sample x. When the attack type i is the same as the attack type label y of attack sample x, i.e., i = y, 1 i＝y The value is 1; when the attack type i is different from the attack type label y of the attack sample x, i.e., i ≠ y, the value is 1. i＝y p is 0; i (x) represents the probability that the attack detection model predicts attack sample x belongs to attack type i; γ represents the balance coefficient; Recall i This represents the recall rate for attack type i.

[0071] In this embodiment, the category-sensitive data of attack types with low recall can be improved, and the computational efficiency can be improved while the implementation complexity can be reduced. This improves the accuracy and efficiency of the determination of the current discrimination loss function, providing an accurate data foundation for the subsequent training of the previous attack detection model.

[0072] S250. Based on the current voting weight combination, the current discriminant loss function, multiple attack samples, and the attack type label corresponding to each attack sample, the previous attack detection model is trained and optimized to obtain the current attack detection model.

[0073] S260. Based on the current attack detection model, identify each attack sample to obtain the judgment type of the corresponding attack sample, and train and optimize the previous attack generation model based on the judgment type and attack type label of each attack sample to obtain the current attack generation model.

[0074] The technical solution of this application embodiment, in response to satisfying a preset training trigger condition, acquires multiple misjudged samples from the previous attack detection model and the attack type label corresponding to each misjudged sample, and determines the key conditions of the previous attack detection model in identifying each misjudged sample, thereby obtaining the conditional constraints of the corresponding misjudged sample. Next, using the previous attack generation model, based on the previous potential attack vector and the conditional constraints of each misjudged sample, it generates attack sample features corresponding to the attack type label, and uses an attack generator to generate attack samples corresponding to the attack sample features based on each attack sample feature. Then, based on multiple misjudged samples, it determines the recall rate of each attack type, and based on each attack type... The recall rate is used to update the previous voting weight combination to obtain the current voting weight combination. The recall rate of each attack type is then used to update the previous discriminant loss function to obtain the current discriminant loss function. This allows for dynamic updates of the previous voting weight combination and discriminant loss function based on the recall rate of each attack type, resulting in higher attention to the voting weights and losses of attack types with low recall rates. This also improves computational efficiency and reduces implementation complexity, thereby increasing the accuracy and efficiency of determining the current voting weight combination and the current discriminant loss function. This provides an accurate data foundation for subsequent training of the previous attack detection model. Then, based on the current voting weight combination and the current discriminant loss function... The attack detection model is trained and optimized using a loss function, multiple attack samples, and attack type labels corresponding to each attack sample to obtain the current attack detection model. Bypass techniques of the attack generation model (i.e., attackers) (such as techniques to modify attack features through deformation, encryption, and obfuscation) can be transformed into evolutionary points for the attack detection model (i.e., defenders), thereby achieving dynamic enhancement and adaptive improvement of the attack detection model. This allows the current attack detection model to accurately identify the attack type corresponding to the attack sample adjusted by the bypass technique, further improving the recognition accuracy of the current attack detection model, especially in complex attack scenarios where attack features are modified through deformation, encryption, and obfuscation. Subsequently, based on the current attack detection model, each attack sample is identified to obtain the corresponding attack sample's judgment type. The previous attack generation model is then trained and optimized based on the judgment type and attack type label of each attack sample to obtain the current attack generation model. Through multiple rounds of alternating training and optimization of the attack detection model and the attack generation model, the type recognition capability of the attack detection model and the attack sample generation capability of the attack generation model can be gradually strengthened, thereby progressively enhancing the attack detection accuracy of the attack detection model in complex attack scenarios.

[0075] Figure 3This is a flowchart illustrating an attack detection method provided in an embodiment of this application. This embodiment can be applied to scenarios requiring attack detection of the actual operational data of electronic devices. The attack detection method provided in this embodiment can be executed by the attack detection device provided in this application, which can be implemented through software and / or hardware. In a specific embodiment, the attack detection device can be integrated into an electronic device, such as a computer. The executing entity of this method can be an electronic device. See also... Figure 3 The attack detection method in this embodiment includes, but is not limited to, the following steps:

[0076] S310. Obtain the actual running data at the current moment.

[0077] Among them, actual operating data refers to the data generated by electronic devices during actual operation.

[0078] S320. Input the actual running data into the current attack detection model to obtain the attack detection results.

[0079] The current attack detection model is obtained by the attack detection model optimization method provided in the aforementioned embodiments. The attack detection result is the result obtained by the current attack detection model after classifying and identifying the actual running data. The attack detection result can be a normal type or a specific attack type.

[0080] S330. Determine whether the attack detection result is within the preset attack type list.

[0081] The preset attack type list is a pre-set list used to store attack types that require warnings. Optionally, high-risk and high-value attack types can be stored in the preset attack type list according to the actual application scenario, so as to provide warnings for attack types that truly need to be focused on.

[0082] Specifically, if the attack detection result is within the preset attack type list, it indicates that an alert is needed for that attack type, and S340 can be executed at this time; otherwise, it indicates that no alert is needed, and S310 can be executed at this time to continue detecting the attack detection result of the actual running data at the next moment.

[0083] S340. When the attack detection result is within the preset attack type list, generate and display the warning information based on the attack detection result.

[0084] The warning message is used to alert administrators that the electronic device is under cyberattack.

[0085] Specifically, when the attack detection result is within the preset attack type list, an early warning message can be generated based on the attack detection result and displayed on the screen of the electronic device. At the same time, an alarm sound can be emitted, or the early warning message can be sent to the user terminal of the administrator so that the user terminal can display the early warning message and prompt the administrator to take timely countermeasures.

[0086] The technical solution of this application embodiment can acquire the actual operating data at the current moment and input the actual operating data into the current attack detection model to obtain the attack detection result. The current attack detection model trained using the attack detection model optimization method provided in the aforementioned embodiment can accurately identify the attack types existing inside electronic devices in complex scenarios, thereby improving the accuracy of the attack detection result determination. Then, when the attack detection result is within the preset attack type list, an early warning information is generated based on the attack detection result and displayed. This can provide early warnings for high-value, high-risk attack types, reducing early warnings for low-value, low-risk attack types. It effectively avoids the problem of important early warning information being overwhelmed by a large number of useless early warnings due to a large number of redundant early warnings. This allows managers to quickly obtain the early warning information that needs to be focused on and take timely countermeasures, thereby ensuring the network security of the device.

[0087] Figure 4 This is a schematic diagram of the attack detection model optimization device provided in an embodiment of this application, with reference to... Figure 4 The attack detection model optimization device may include:

[0088] The determination module 410 is used to obtain multiple misjudged samples of the previous attack detection model and the attack type label corresponding to each misjudged sample in response to the satisfaction of the preset training trigger condition, and to determine the key conditions of the previous attack detection model when identifying each misjudged sample, so as to obtain the condition constraints of the corresponding misjudged sample.

[0089] The generation module 420 is used to generate attack sample features corresponding to the attack type label by using the previous attack generation model, based on the previous potential attack vector and the conditional constraints of each misjudged sample, and to use the attacker to generate attack samples corresponding to the attack sample features based on each attack sample feature.

[0090] The first training module 430 is used to update the previous voting weight combination and the previous discriminant loss function of the previous attack detection model based on multiple misjudged samples, thereby obtaining the current voting weight combination and the current discriminant loss function. Based on the current voting weight combination, the current discriminant loss function, multiple attack samples and the attack type label corresponding to each attack sample, the previous attack detection model is trained and optimized to obtain the current attack detection model.

[0091] In one embodiment, the first training module 430 updates the previous voting weight combination and the previous discriminant loss function of the previous attack detection model based on multiple misjudged samples, thereby obtaining the current voting weight combination and the current discriminant loss function. This includes: determining the recall rate of each attack type based on multiple misjudged samples; updating the previous voting weight combination based on the recall rate of each attack type to obtain the current voting weight combination; and updating the previous discriminant loss function based on the recall rate of each attack type to obtain the current discriminant loss function.

[0092] In one embodiment, the first training module 430 updates the previous voting weight combination based on the recall rate of each attack type to obtain the current voting weight combination, including: obtaining the preset basic weights of each attack type; determining the current voting weight of the corresponding attack type based on the previous discriminative learning rate of the previous attack detection model, the recall rate of each attack type, and the preset basic weights; and updating the previous voting weight combination based on the current voting weights of each attack type to obtain the current voting weight combination.

[0093] In one embodiment, the first training module 430 updates the previous discriminative loss function based on the recall rate of each attack type to obtain the current discriminative loss function, including: determining the current category sensitive data of the corresponding attack type based on the previous balance coefficient of the previous attack detection model and the recall rate of each attack type; and updating the previous category sensitive data of the corresponding attack type in the previous discriminative loss function based on the current category sensitive data of each attack type to obtain the current discriminative loss function.

[0094] In one embodiment, the determining module 410 determines the key conditions for the previous attack detection model in identifying each misclassified sample, including:

[0095] For the current misjudged sample among multiple misjudged samples, determine the discrimination condition path of the previous attack detection model when identifying the current misjudged sample, and determine the condition corresponding to the node with the minimum information entropy in the discrimination condition path as the key condition of the current misjudged sample.

[0096] In one embodiment, the attack detection model optimization device further includes a second training module, which is used to: identify each attack sample based on the current attack detection model to obtain the judgment type of the corresponding attack sample, and train and optimize the previous attack generation model based on the judgment type and attack type label of each attack sample to obtain the current attack generation model.

[0097] In one embodiment, the preset training triggering condition in the determination module 410 includes the number of newly added misjudged samples being greater than a preset number threshold.

[0098] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is merely an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the functional modules described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0099] The attack detection model optimization device provided in this embodiment can be applied to the attack detection model optimization method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0100] Figure 5 This is a schematic diagram of the attack detection device provided in an embodiment of this application, referring to... Figure 5 The attack detection device may include:

[0101] The acquisition module 510 is used to acquire the actual running data at the current moment;

[0102] The attack detection module 520 is used to input actual running data into the current attack detection model obtained by the attack detection model optimization method of any embodiment of this application, and obtain the attack detection result.

[0103] The early warning module 530 is used to generate and display early warning information based on the attack detection results when the attack detection results are within the preset attack type list.

[0104] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is merely an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the functional modules described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0105] The attack detection device provided in this embodiment can be applied to the attack detection methods provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0106] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 6 A block diagram is shown of an exemplary electronic device 11 suitable for implementing embodiments of the present application. Figure 6 The electronic device 11 shown is merely an example and should not impose any limitations on the functionality and scope of use of this embodiment.

[0107] like Figure 6As shown, the electronic device 11 is represented in the form of a general-purpose computing electronic device. The components of the electronic device 11 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0108] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0109] Electronic device 11 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by electronic device 11, including volatile and non-volatile media, removable and non-removable media.

[0110] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Electronic device 11 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (… Figure 6 Not shown; usually referred to as a "hard drive"). Although Figure 6 As not shown, disk drives for reading and writing to removable non-volatile disks (e.g., "floppy disks") and optical disc drives for reading and writing to removable non-volatile optical discs (e.g., CD-ROMs, DVD-ROMs, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application.

[0111] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of this application.

[0112] Electronic device 11 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with electronic device 11, and / or with any device that enables electronic device 11 to communicate with one or more other computing devices (e.g., network interface card and modem, etc.). This communication can be performed via input / output (I / O) interface 22. Furthermore, electronic device 11 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20.

[0113] like Figure 6 As shown, network adapter 20 communicates with other modules of electronic device 11 via bus 18. It should be understood that, although... Figure 6 As not shown, other hardware and / or software modules may be used in conjunction with electronic device 11, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0114] The processing unit 16 executes various functional applications and page displays by running programs stored in the system memory 28, such as implementing an attack detection model optimization method provided in any embodiment of this application, or implementing an attack detection method provided in any embodiment of this application.

[0115] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements, for example, an attack detection model optimization method provided in any embodiment of this application, or, when executed by a processor, implements, for example, an attack detection method provided in any embodiment of this application.

[0116] The computer storage medium of this embodiment can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0117] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0118] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0119] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0120] Those skilled in the art will understand that the modules or steps described above in this application can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, which can then be stored in a storage device for execution by a computing device. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0121] Furthermore, the acquisition, storage, use, and processing of data in this application's technical solution all comply with relevant laws and regulations.

[0122] Note that the above are merely preferred embodiments and the technical principles employed in this application. Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments. Many other equivalent embodiments may be included without departing from the inventive concept of this application, and the scope of this application is determined by the scope of the appended claims.

Claims

1. An attack detection model optimization method, characterized in that, The method includes: In response to the satisfaction of the preset training triggering conditions, multiple misjudged samples of the previous attack detection model and the attack type label corresponding to each misjudged sample are obtained, and the key conditions of the previous attack detection model in identifying each misjudged sample are determined, so as to obtain the condition constraints of the corresponding misjudged sample. Using the previous attack generation model, based on the previous potential attack vector and the conditional constraints of each misjudged sample, attack sample features with corresponding attack type labels are generated, and the attacker generates attack samples with corresponding attack sample features based on each attack sample feature. The previous attack detection model is updated based on multiple misjudged samples, which yields the current voting weight combination and the current discrimination loss function. The previous attack detection model is then trained and optimized based on the current voting weight combination, the current discrimination loss function, multiple attack samples, and the attack type label corresponding to each attack sample, to obtain the current attack detection model.

2. The attack detection model optimization method according to claim 1, characterized in that, Based on multiple misclassified samples, the previous voting weight combination and the previous discriminant loss function of the previous attack detection model are updated to obtain the current voting weight combination and the current discriminant loss function, including: The recall rate for each attack type was determined based on multiple false positive samples. The previous voting weight combination is updated based on the recall rate of each attack type to obtain the current voting weight combination. The previous discriminant loss function is also updated based on the recall rate of each attack type to obtain the current discriminant loss function.

3. The attack detection model optimization method according to claim 2, characterized in that, The previous voting weight combination is updated based on the recall rate of each attack type to obtain the current voting weight combination, including: Obtain the preset base weights for each attack type; Based on the previous discrimination learning rate of the previous attack detection model, the recall rate of each attack type, and the preset basic weights, the current voting weight of the corresponding attack type is determined. The previous voting weight combination is updated based on the current voting weight of each attack type to obtain the current voting weight combination.

4. The attack detection model optimization method according to claim 2, characterized in that, The previous discriminative loss function is updated based on the recall rate of each attack type to obtain the current discriminative loss function, which includes: Based on the previous balance coefficient of the previous attack detection model and the recall rate of each attack type, determine the current category of sensitive data for the corresponding attack type; Based on the current category sensitive data of each attack type, the previous category sensitive data of the corresponding attack type in the previous discriminative loss function is updated to obtain the current discriminative loss function.

5. The attack detection model optimization method according to claim 1, characterized in that, The key conditions for the previous attack detection model to identify each false positive sample were determined, including: For the current misjudged sample among multiple misjudged samples, determine the discrimination condition path of the previous attack detection model when identifying the current misjudged sample, and determine the condition corresponding to the node with the minimum information entropy in the discrimination condition path as the key condition of the current misjudged sample.

6. The attack detection model optimization method according to claim 1, characterized in that, The method further includes: Each attack sample is identified based on the current attack detection model to obtain the corresponding attack sample's determination type. The previous attack generation model is then trained and optimized based on the determination type and attack type label of each attack sample to obtain the current attack generation model.

7. The attack detection model optimization method according to claim 1, characterized in that, The preset training trigger conditions include the number of newly added misclassified samples exceeding a preset threshold.

8. An attack detection method, characterized in that, The method includes: Obtain the actual operating data at the current moment; The actual operating data is input into the current attack detection model obtained by the attack detection model optimization method described in any one of claims 1 to 7 to obtain the attack detection result. When the attack detection result is within the preset attack type list, an early warning message is generated based on the attack detection result and displayed.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the attack detection model optimization method according to any one of claims 1 to 7, or to enable the at least one processor to perform the attack detection method according to claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the attack detection model optimization method as described in any one of claims 1 to 7, or when the program is executed by the processor, it implements the attack detection method as described in claim 8.

Images