Escape attack intrusion detection method and device and storage medium

Abstract

The invention relates to an escape attack intrusion detection method and device and a storage medium. The method comprises the following steps: arranging monitoring equipment; obtaining original seismic wave data; acquiring monitoring data; acquiring monitoring seismic wave data according to the monitoring data; and comparing the original seismic wave data with the monitored seismic wave data to judge a fireflooding ignition success result. According to the escape attack intrusion detection method and device and the storage medium provided by the invention, based on the GAN thought, the target network F capable of effectively detecting the escape attack behavior is mainly trained while the G network and the D network are trained; therefore, an attacker bypasses the intrusion detection logic based on machine learning by adopting escape attacks to become an impossible event.

Description

Technical Field

[0001] The present application relates to the field of escape attack intrusion detection, and in particular to an escape attack intrusion detection method, device and storage medium. Background Art

[0002] With the continuous application of machine learning technology in the field of intrusion detection, attackers have gradually developed various attack methods that can deceive machine learning-based intrusion detection models for escape attack detection in industrial control networks. Escape attack is one of the most effective deception methods. It refers to an attack in which the attacker constructs specific input samples to deceive the target system without changing the target machine learning system.

[0003] The mainstream intrusion detection technology research for evasion attacks mainly focuses on the following directions:

[0004] Direction 1: Defense distillation: This method adopts the idea of ​​knowledge transfer and transfers the information learned by the original classification model to a small network model, thereby achieving gradient masking. Defense distillation can effectively resist some gradient-based escape attacks, but studies have shown that this detection method fails in the case of unknown model functions or black-box attacks, such as when using the C&W method to launch attacks.

[0005] Direction 2: Regularization: This method refers to penalizing the output of the objective function of the classification model during the training process, thereby reducing the sensitivity of the model's output to the input. The detection method of deep compression network is similar to the idea of ​​this method. It adds a smooth penalty term by using compressed autoencoders during the training process to make the model's output smoother. This method can significantly improve the robustness of the detector, but may make the detector effect worse to a certain extent.

[0006] Direction 3: Defending against global perturbations: This method adds an additional processing layer before the classification model and trains the layer so that it has the ability to restore adversarial samples of global perturbations. However, this method can only resist global perturbations against the model and cannot detect adversarial samples generated by other methods.

[0007] Direction 4: Detection based on generative adversarial networks: This method uses generative adversarial networks to enhance the robustness of the detector. It includes two machine learning models, one as a detector and one for generating adversarial samples, and then the detector is continuously trained through the game between the two models. This method can effectively resist adversarial attacks, but if the generative adversarial network itself is not properly trained and challenged, the model will be affected by the original input samples and adversarial samples.

[0008] Direction 5: Detecting adversarial samples: This type of defense method only needs to identify whether the input sample is an adversarial sample. Currently, researchers have proposed a variety of detection methods, such as statistical-based detection, local intrinsic dimension-based detection, and feature squeezing. However, this type of method generally does not provide a response solution after detecting an adversarial sample, and it will also reduce the accuracy of the detector.

[0009] Direction 6: Adversarial training: This method is to put the adversarial samples generated by each method into the original model for training, and improve the robustness of the classification model by modifying the training sample set. In addition to classic adversarial training, there are many optimized adversarial training methods, such as integrated adversarial training and cascaded adversarial training, which have achieved good detection results. However, this method can only add specific types of adversarial samples generated by known attack methods and lacks generalization ability. In addition, when the detector data set is large, the training cost of this method is high.

[0010] Direction 7: Data compression: This method is to reduce the impact of adversarial perturbations of adversarial samples on the classification model by modifying the sample set in the test phase. For example, multiple traffic input samples are compressed to form an input sample before the detector makes a judgment. The advantage of data compression is that it does not require model modification, has a small amount of calculation and is simple to implement, but it will have a certain impact on the accuracy of the detector. Summary of the invention

[0011] In order to solve the above technical problems or at least partially solve the above technical problems, the present application provides an escape attack intrusion detection method, device and storage medium.

[0012] In a first aspect, the present application provides an evasion attack intrusion detection method, the method comprising the steps of:

[0013] Obtain network behavior characteristics;

[0014] Performing feature processing on the network behavior features;

[0015] Construct evasion attack intrusion detection parameters;

[0016] Using the processing results to train the evasion attack intrusion detection parameters;

[0017] Output the escape attack intrusion detection model based on the training results.

[0018] Preferably, the obtaining of network behavior characteristics comprises the steps of:

[0019] Extract network behavior characteristics in actual industrial control networks;

[0020] Aggregating the network behavior features to form a network behavior data set;

[0021] Data annotation processing is performed on each network behavior in the network behavior dataset.

[0022] Preferably, the feature processing of the network behavior feature comprises the steps of:

[0023] Acquire a network behavior data set of the network behavior characteristics;

[0024] Clearing unreasonable value data samples in the network behavior data set;

[0025] Clearing the erroneous value data samples in the network behavior data set;

[0026] Clear the repeated value data samples in the network behavior data set.

[0027] Preferably, the step of constructing the escape attack intrusion detection parameters comprises the following steps:

[0028] Build the AdvGAN model;

[0029] Construct the loss function of the GAN network;

[0030] Construct the loss function of the F network;

[0031] Construct an early stopping mechanism.

[0032] Preferably, the using the processing results to train the escape attack intrusion detection parameters comprises the steps of:

[0033] Obtaining the AdvGAN model in the escape attack intrusion detection parameters;

[0034] Obtaining feature processing results of the network behavior features;

[0035] The AdvGAN model is trained using the feature processing results.

[0036] Preferably, the step of outputting the escape attack intrusion detection model according to the training results comprises the following steps:

[0037] Obtaining an early stopping mechanism in the escape attack intrusion detection parameter;

[0038] Get the trained AdvGAN model;

[0039] Using the early stopping mechanism to determine whether the AdvGAN model has converged;

[0040] If yes, output the escape attack intrusion detection model in the trained F network;

[0041] If not, return to the method of using the processing result to train the escape attack intrusion detection parameters.

[0042] Preferably, the method further comprises the steps of:

[0043] The evasion attack intrusion detection model is deployed on the intrusion detection device in the actual industrial control network environment.

[0044] In a second aspect, the present application provides an escape attack intrusion detection device, comprising:

[0045] A feature acquisition module is used to acquire network behavior features;

[0046] A feature processing module, used for performing feature processing on the network behavior features;

[0047] Parameter construction module, used to construct escape attack intrusion detection parameters;

[0048] A parameter training module, used for training the escape attack intrusion detection parameters using the processing results;

[0049] The model output module is used to output the escape attack intrusion detection model according to the training results.

[0050] In a third aspect, an air conditioner control device is provided, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0051] Memory, used to store computer programs;

[0052] The processor is used to implement the steps of the escape attack intrusion detection method described in any one of the embodiments of the first aspect when executing the program stored in the memory.

[0053] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the escape attack intrusion detection method as described in any embodiment of the first aspect are implemented.

[0054] The above technical solution provided by the embodiment of the present application has the following advantages compared with the prior art:

[0055] The present application provides an escape attack intrusion detection method, device and storage medium based on the GAN concept. While training the G network and the D network, the focus is on training a target network F that can effectively detect escape attack behaviors, making it impossible for an attacker to bypass the intrusion detection logic based on machine learning by taking escape attacks. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0058] Figure 1 It is a flow chart of an evasion attack intrusion detection method provided by an embodiment of the present invention;

[0059] Figure 2 It is a structural diagram of an escape attack intrusion detection device provided by an embodiment of the present invention;

[0060] Figure 3 It is a structural schematic diagram of an electronic device provided by the present invention;

[0061] Figure 4 It is a structural schematic diagram of a non-transitory computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0063] Figure 1 A flowchart of an evasion attack intrusion detection method provided in an embodiment of the present application.

[0064] The present application provides an evasion attack intrusion detection method, the method comprising the steps of:

[0065] S1: Obtain network behavior characteristics;

[0066] In the embodiment of the present application, the obtaining of network behavior characteristics comprises the steps of:

[0067] Extract network behavior characteristics in actual industrial control networks;

[0068] Aggregating the network behavior features to form a network behavior data set;

[0069] Data annotation processing is performed on each network behavior in the network behavior dataset.

[0070] Specifically, network behavior feature extraction is carried out in the actual industrial control network to form a network behavior data set, and data annotation processing is performed on each network behavior in the data set.

[0071] S2: performing feature processing on the network behavior features;

[0072] In an embodiment of the present application, the feature processing of the network behavior feature includes the steps of:

[0073] Acquire a network behavior data set of the network behavior characteristics;

[0074] Clearing unreasonable value data samples in the network behavior data set;

[0075] Clearing the erroneous value data samples in the network behavior data set;

[0076] Clear the repeated value data samples in the network behavior data set.

[0077] Specifically, feature engineering processing is performed on the data set generated in step S1, with a focus on removing data samples such as unreasonable values, erroneous values, and repeated values ​​in the data set.

[0078] S3: Construct escape attack intrusion detection parameters;

[0079] In an embodiment of the present application, the step of constructing escape attack intrusion detection parameters includes the following steps:

[0080] Build the AdvGAN model;

[0081] Construct the loss function of the GAN network;

[0082] Construct the loss function of the F network;

[0083] Construct an early stopping mechanism.

[0084] Specifically, the loss functions of the AdvGAN model, the GAN network and the F network are constructed, wherein the G network represents the generation network, which is used to construct the intrusion sample data with interference information; the D network represents the discrimination network, which is used to distinguish between intrusion behavior and normal behavior; the IDS network, that is, the F network described in the present invention (hereinafter referred to as the F network), is used to detect intrusion behavior with escape attack characteristics.

[0085] S4: Using the processing results to train the evasion attack intrusion detection parameters;

[0086] In an embodiment of the present application, the step of using the processing results to train the escape attack intrusion detection parameters includes the following steps:

[0087] Obtaining the AdvGAN model in the escape attack intrusion detection parameters;

[0088] Obtaining feature processing results of the network behavior features;

[0089] The AdvGAN model is trained using the feature processing results.

[0090] Specifically, the constructed AdvGAN model is trained using an early stopping mechanism.

[0091] S5: Output the escape attack intrusion detection model according to the training results.

[0092] In an embodiment of the present application, outputting the escape attack intrusion detection model according to the training results includes the steps of:

[0093] Obtaining an early stopping mechanism in the escape attack intrusion detection parameter;

[0094] Get the trained AdvGAN model;

[0095] Using the early stopping mechanism to determine whether the AdvGAN model has converged;

[0096] If yes, output the escape attack intrusion detection model in the trained F network;

[0097] If not, return to the method of using the processing result to train the escape attack intrusion detection parameters.

[0098] Specifically, the early stopping mechanism is used to determine whether the AdvGAN model converges. If converged, it means that the value of L is no longer significantly lower. At this time, the escape attack intrusion detection model in the trained F network is output, otherwise, step S4 is continued to carry out cyclic training.

[0099] In an embodiment of the present application, the step of deploying the escape attack intrusion detection model on an intrusion detection device in an actual industrial control network environment is also included.

[0100] Specifically, when deploying the evasion attack intrusion detection model on the intrusion detection device in the actual industrial control network environment, the following steps can be taken to make it effective:

[0101] Step 1: Deploy the evasion attack intrusion detection model on the intrusion detection device in the actual industrial control network environment;

[0102] Step 2: Carry out feature extraction to extract characteristic data of network behavior, which requires a certain amount of data and necessary feature engineering;

[0103] Step 3: Use the deployed evasion attack intrusion detection model to detect the network behavior data after feature extraction;

[0104] Step 4: If the detection results are collected, the collected detection results are regularly fed back to the training phase of the escape attack intrusion detection model to carry out iterative upgrades of the escape attack intrusion attack monitoring model; otherwise, directly proceed to step 5 for processing;

[0105] Step 5: Automatically generate corresponding processing logs based on the detection results, and automatically generate interception strategies as needed to intercept corresponding intrusion attack behaviors.

[0106] like Figure 2 , the present application provides an escape attack intrusion detection device, comprising:

[0107] A feature acquisition module 10 is used to acquire network behavior features;

[0108] A feature processing module 20, used for performing feature processing on the network behavior feature;

[0109] A parameter construction module 30, used to construct escape attack intrusion detection parameters;

[0110] A parameter training module 40, used for training the escape attack intrusion detection parameters using the processing results;

[0111] The model output module 50 is used to output the escape attack intrusion detection model according to the training results.

[0112] An escape attack intrusion detection device provided in the present application can execute an escape attack intrusion detection method provided in the above steps.

[0113] It should be understood that the above specific embodiments of the present invention are only used to illustrate or explain the principles of the present invention, and do not constitute a limitation of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. In addition, the appended claims of the present invention are intended to cover all changes and modifications that fall within the scope and boundaries of the appended claims, or the equivalent forms of such scope and boundaries.

[0114] Reference below Figure 3 , which shows a schematic diagram of the structure of an electronic device 100 suitable for implementing the embodiment of the present disclosure. The electronic device in the embodiment of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0115] like Figure 3 As shown, the electronic device 100 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 102 or a program loaded from a storage device 108 to a random access memory (RAM) 103. In the RAM 103, various programs and data required for the operation of the electronic device 100 are also stored. The processing device 101, the ROM 102, and the RAM 103 are connected to each other via a bus 104. An input / output (I / O) interface 105 is also connected to the bus 104.

[0116] Typically, the following devices may be connected to the I / O interface 105: an input device 106 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 107 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 108 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 109. The communication device 109 may allow the electronic device 100 to communicate with other devices wirelessly or by wire to exchange data. Although the electronic device 100 having various devices is shown in the figure, it should be understood that it is not required to implement or have all the devices shown. More or fewer devices may be implemented or have alternatively.

[0117] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 109, or installed from the storage device 108, or installed from the ROM 102. When the computer program is executed by the processing device 101, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0118] Reference below Figure 4 , which shows a structural schematic diagram of a computer-readable storage medium suitable for implementing an embodiment of the present disclosure, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it can implement any of the evasion attack intrusion detection methods described above.

[0119] The present application provides an escape attack intrusion detection method, device and storage medium based on the GAN concept. While training the G network and the D network, the focus is on training a target network F that can effectively detect escape attack behaviors, making it impossible for an attacker to bypass the intrusion detection logic based on machine learning by taking escape attacks.

[0120] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0121] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An evasion attack intrusion detection method, characterized in that: The method comprises the steps of: Obtain network behavior characteristics; Performing feature processing on the network behavior features; Construct evasion attack intrusion detection parameters; Using the processing results to train the evasion attack intrusion detection parameters; Output the escape attack intrusion detection model based on the training results.

2. The evasion attack intrusion detection method according to claim 1, characterized in that: The obtaining of network behavior characteristics comprises the steps of: Extract network behavior characteristics in actual industrial control networks; Aggregating the network behavior features to form a network behavior data set; Data annotation processing is performed on each network behavior in the network behavior dataset.

3. The evasion attack intrusion detection method according to claim 1, characterized in that: The feature processing of the network behavior feature comprises the steps of: Acquire a network behavior data set of the network behavior characteristics; Clearing unreasonable value data samples in the network behavior data set; Clearing the erroneous value data samples in the network behavior data set; Clear the repeated value data samples in the network behavior data set.

4. The evasion attack intrusion detection method according to claim 1, characterized in that: The construction of escape attack intrusion detection parameters comprises the steps of: Build the AdvGAN model; Construct the loss function of the GAN network; Construct the loss function of the F network; Construct an early stopping mechanism.

5. The evasion attack intrusion detection method according to claim 1, characterized in that: The method of using the processing results to train the escape attack intrusion detection parameters comprises the steps of: Obtaining the AdvGAN model in the escape attack intrusion detection parameters; Obtaining feature processing results of the network behavior features; The AdvGAN model is trained using the feature processing results.

6. The evasion attack intrusion detection method according to claim 1, characterized in that: Outputting the escape attack intrusion detection model according to the training results comprises the following steps: Obtaining an early stopping mechanism in the escape attack intrusion detection parameter; Get the trained AdvGAN model; Using the early stopping mechanism to determine whether the AdvGAN model has converged; If yes, output the escape attack intrusion detection model in the trained F network; If not, return to the method of using the processing result to train the escape attack intrusion detection parameters.

7. The evasion attack intrusion detection method according to claim 1, characterized in that: Also includes the steps: The evasion attack intrusion detection model is deployed on the intrusion detection device in the actual industrial control network environment.

8. An escape attack intrusion detection device, characterized in that: include: A feature acquisition module is used to acquire network behavior features; A feature processing module, used for performing feature processing on the network behavior features; Parameter construction module, used to construct escape attack intrusion detection parameters; A parameter training module, used for training the escape attack intrusion detection parameters using the processing results; The model output module is used to output the escape attack intrusion detection model according to the training results.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor is used to implement the steps of the evasion attack intrusion detection method described in any one of claims 1 to 7 when executing the program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the evasion attack intrusion detection method according to any one of claims 1 to 7 are implemented.

Images