A face recognition attack processing method, device and equipment and a storage medium

Abstract

This application discloses a method, apparatus, device, and storage medium for processing facial recognition attacks. It calculates facial image similarity values ​​by using a pre-set facial similarity calculation model to analyze the acquired facial metadata and post-event facial record data corresponding to the face. Then, based on the post-event facial record data and facial image similarity values, it constructs a connected graph of facial recognition post-event record similarity, with nodes containing state information records corresponding to the face and edges containing facial image similarity values. Finally, it inputs the connected graph data into a pre-set deep graph neural network model for post-event screening and classification calculation of facial recognition attacks. This solves the technical problem currently faced in post-event screening and defense against facial recognition attacks using masks, facial stickers, or other non-human-related attacks, where many non-human attack cases are not accurately identified.

Description

Technical Field

[0001] This application relates to the field of financial technology, and in particular to a method, apparatus, device and storage medium for handling facial recognition attacks. Background Technology

[0002] Due to the rapid development of computer science and electronic technology, facial recognition has become the world's second-largest biometric authentication method, second only to fingerprint recognition, in terms of market share. More and more fintech products are incorporating facial recognition functionality. It's required in many business scenarios, such as logging into bank apps, transferring funds, applying for loan products, and purchasing wealth management products. Simultaneously, some remote and online intelligent services also require facial recognition to verify customer identity. Facial recognition can verify that the customer applying for the service matches the person in the photograph, preventing unauthorized applications. It can also verify that the customer is a living person, preventing the use of photos or pre-recorded videos for facial recognition during application processes.

[0003] However, there are currently malicious attacks that use customers' photos or "image animation technology" to "reproduce" customer photos, attempting to bypass the bank's facial recognition function to steal customer account funds. Therefore, technologies to prevent facial recognition attacks have emerged. The currently used "post-attack screening of facial recognition attacks" technology performs model calculations and screening classifications on the massive records of facial photos or videos of customers after facial recognition, confirming whether there are cases of black market attacks or even breaches of the bank's facial recognition function in the customer's historical post-attack facial recognition records.

[0004] However, when it comes to post-event screening and defense against facial recognition attacks using "non-person" methods such as wearing masks or face stickers, there are many technical issues where "non-person" attack cases are not accurately identified. Summary of the Invention

[0005] This application provides a method, apparatus, device, and storage medium for handling facial recognition attacks, which solves the technical problem that currently used facial recognition attacks involving non-human entities such as those wearing masks or facial stickers cannot be accurately identified in post-event screening and defense.

[0006] In view of this, the first aspect of this application provides a method for handling face recognition attacks, the method comprising:

[0007] S1. Calculate the face image similarity value by using a pre-set face similarity calculation model to obtain the face metadata and post-face record data corresponding to the face.

[0008] S2. Based on the face post-record data and face image similarity values, construct a connected graph of face recognition post-record similarity data, with state information records corresponding to the face as node data and face image similarity values ​​as edge data.

[0009] S3. Input the similarity connected graph data of the face recognition post-event record into the preset deep graph neural network model to perform face recognition attack post-event screening and classification calculation.

[0010] Preferably, the method further includes the following steps before step S1:

[0011] S4. Obtain the face metadata corresponding to the face, as well as the face post-recording data corresponding to the preset business scenario.

[0012] Preferably, step S1 specifically includes:

[0013] S11. Extract face images from face metadata and post-face recording data to obtain at least two face recognition image data;

[0014] S12. Pair the face recognition image data in pairs, and perform face key point recognition, face pose correction, and face feature extraction processing according to the landmark algorithm.

[0015] S13. The facial recognition image data after facial key point recognition, facial pose correction, and facial feature extraction is processed by a pre-set facial similarity calculation model to calculate the facial image similarity value.

[0016] Preferably, step S3 specifically includes:

[0017] S31. Input the connected graph data of the similarity record after face recognition into a dynamic graph attention network model to suppress the oversmoothing phenomenon;

[0018] S32. By using a dynamic graph attention network model to suppress oversmoothing, the information transfer and aggregation calculation of neighbor node data between each pair of neighbor nodes in the edge direction of the node data corresponding to the edge data is performed on the similarity connected graph data of face recognition after the fact.

[0019] S33. Input the data information transmission and aggregation calculation results of neighboring nodes into the screening and classification layer network model to perform post-face recognition attack screening and classification calculation.

[0020] Preferably, the pre-set face similarity calculation model includes:

[0021]

[0022]

[0023] Among them, i o / i1 represents two paired face recognition images; CNN n It is a multi-layer convolutional neural network; These are the data features extracted from two paired face recognition images; similarity dual This refers to the similarity score between images of people's faces.

[0024] Preferably, the dynamic graph attention network model that suppresses oversmoothing includes:

[0025]

[0026] Among them, v j For the current node v i The data of any pair of neighboring nodes, and v k For the current node v i E represents the data of all pairs of n-1 neighbor nodes. i，j / E i，k Let E be the feature E of the edge data connecting pairs of neighboring node data. T Θ represents the affine transformation operation of the fully connected layer network, LeakyReLU is the activation function, and exp represents the exponential operation with the natural number e as the base.

[0027] Preferably, the face metadata includes static face recognition associated data;

[0028] Post-event facial recognition data includes dynamic scene facial recognition data associated with pre-defined business scenarios.

[0029] A second aspect of this application provides a face recognition attack processing apparatus, the apparatus comprising:

[0030] The similarity calculation unit is used to calculate the face image similarity value based on the obtained face metadata and post-face record data corresponding to the face through a pre-set face similarity calculation model.

[0031] The similarity connected graph unit is used to construct a face recognition post-record similarity connected graph data based on face post-record data and face image similarity values. The graph data consists of state information records corresponding to the face as node data and face image similarity values ​​as edge data.

[0032] The screening and classification calculation unit is used to input the similarity connected graph data of the face recognition post-event record into a preset deep graph neural network model to perform face recognition attack post-event screening and classification calculation.

[0033] A third aspect of this application provides a face recognition attack processing device, the device including a processor and a memory:

[0034] The memory is used to store program code and transfer the program code to the processor;

[0035] The processor is used to execute the steps of the face recognition attack handling method described in the first aspect above, according to the instructions in the program code.

[0036] The fourth aspect of this application provides a computer-readable storage medium for storing program code for performing the steps of the face recognition attack processing method of the first aspect described above.

[0037] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0038] This application provides a method for handling face recognition attacks. It calculates the face image similarity value by using a pre-set face similarity calculation model to analyze the acquired face metadata and post-event face record data. Then, based on the post-event face record data and face image similarity values, it constructs a connected graph of face recognition post-event record similarity, with nodes containing state information records corresponding to the face and edges containing face image similarity values. Finally, it inputs the connected graph data into a pre-set deep graph neural network model for post-event screening and classification of face recognition attacks. This method solves the technical problem currently faced in post-event screening and defense against non-human face recognition attacks such as those using masks or face stickers, where many non-human attack cases are not accurately identified.

[0039] Furthermore, this application employs a dynamic graph attention network model that suppresses oversmoothing, which can accurately screen, classify, and identify post-event facial recognition records that are either "not the person in question" or "electronic screen photocopying" type black market facial recognition attacks. Attached Figure Description

[0040] Figure 1 This is a flowchart of one embodiment of a face recognition attack handling method according to the present application;

[0041] Figure 2 This is a schematic diagram of the connected graph data of the face recognition post-recognition similarity recording in the embodiments of this application;

[0042] Figure 3 This is a schematic diagram of the structure of a face recognition attack device according to an embodiment of this application;

[0043] Figure 4 This is a schematic diagram of the structure of a face recognition attack device according to an embodiment of this application. Detailed Implementation

[0044] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0045] This application designs a method, apparatus, device, and storage medium for handling facial recognition attacks, which solves the technical problem that currently used facial recognition attacks involving non-human entities such as those wearing masks or face stickers cannot be accurately identified in post-event screening and defense.

[0046] For easier understanding, please refer to Figure 1 , Figure 1 This is a flowchart of a face recognition attack handling method according to an embodiment of this application, such as... Figure 1 As shown, specifically:

[0047] S1. Calculate the face image similarity value by using a pre-set face similarity calculation model to obtain the face metadata and post-face record data corresponding to the face.

[0048] It should be noted that before step S1, the process also includes obtaining the facial metadata corresponding to the face, as well as the facial post-recording data corresponding to the pre-set business scenario. Among them, the "metadata record" of facial recognition is to obtain data such as "customer's mobile phone model", "business scenario code when customer performs bank facial recognition", and "similarity score between the customer's current facial recognition capture photo and the photo kept by the bank or public security" when the customer performs facial recognition.

[0049] The "post-event image record" data of facial recognition refers to the image data captured by the facial recognition system when a customer undergoes facial recognition in a certain business scenario at the bank.

[0050] Furthermore, the specific steps in this embodiment are as follows:

[0051] S11. Extract face images from face metadata and post-face recording data to obtain at least two face recognition image data;

[0052] S12. Pair the face recognition image data in pairs, and perform face key point recognition, face pose correction, and face feature extraction processing according to the landmark algorithm.

[0053] S13. The facial recognition image data after facial key point recognition, facial pose correction, and facial feature extraction is processed by a pre-set facial similarity calculation model to calculate the facial image similarity value.

[0054] It should be noted that before constructing the similarity connected graph data of the post-facial recognition record, it is necessary to use an "image similarity calculation model" to calculate the "face image similarity value" between all "facial recognition image data" in each customer's "post-facial recognition image record".

[0055] The "landmark facial landmark recognition" module identifies the main landmarks of the "face region" in the facial recognition image. Next, the "facial pose correction" module corrects the angle of the "face region" in the facial recognition image based on the identified landmarks. Then, the "facial feature extraction" module extracts facial region data features from the face recognition image after angle correction. Finally, the facial region data features extracted by the "facial feature extraction" module are input into the "facial similarity calculation model," which outputs the "facial image similarity value" between the two "facial recognition image data." The "facial similarity calculation model" used here is shown in the following formula:

[0056]

[0057]

[0058] In the formula above, i o / i1 represents two paired "face recognition image data"; CNN n This represents a multi-layer convolutional neural network used to extract data features from the "face region". This refers to the data features extracted from the "face region" in two paired "face recognition image data" images; similarity dual This represents the "face image similarity value" between the two "face recognition image data" that are finally calculated.

[0059] S2. Based on the face post-record data and face image similarity values, construct a connected graph of face recognition post-record similarity data, with state information records corresponding to the face as node data and face image similarity values ​​as edge data.

[0060] It should be noted that, as Figure 2 As shown, it contains two parts of data: "nodes" and "edges," as shown in the following formula:

[0061]

[0062]

[0063]

[0064] In the formula above, G simi This represents "post-record similarity connected graph data of facial recognition"; n represents G. simi The number of "node" data in G; m represents G simi The number of "edge" data in the data.

[0065] V n For this similarity connected graph data, all node data, namely "post-facial recognition record data", each node data contains sub-data such as "customer's mobile phone model" P, "customer's business scenario code when performing bank facial recognition" S, and "similarity score between the customer's current facial recognition capture photo and the photo kept by the bank or public security" B. m For this similarity connected graph data, each edge data contains a subdata item called "similarity" representing the pairwise similarity values ​​between images after face recognition. dual .

[0066] In constructing the "face recognition post-recording similarity connected graph data" G simi In this context, if there are n "node" data points, then each "node" data point will have n-1 "neighbor" data points, and pairs of "neighbor" data points will be connected by an "edge" E. m Connected.

[0067] S3. Input the similarity connected graph data of the face recognition post-event record into the preset deep graph neural network model to perform face recognition attack post-event screening and classification calculation.

[0068] It should be noted that the specific steps in this embodiment include:

[0069] S31. Input the connected graph data of the similarity record after face recognition into a dynamic graph attention network model to suppress the oversmoothing phenomenon;

[0070] S32. By using a dynamic graph attention network model to suppress oversmoothing, the information transfer and aggregation calculation of neighbor node data between each pair of neighbor nodes in the edge direction of the node data corresponding to the edge data is performed on the similarity connected graph data of face recognition after the fact.

[0071] S33. Input the data information transmission and aggregation calculation results of neighboring nodes into the screening and classification layer network model to perform post-face recognition attack screening and classification calculation.

[0072] Furthermore, the screening categories in this embodiment are as follows:

[0073] After the "face recognition post-event record similarity connected graph data" is constructed, it can be input into the constructed "deep graph neural network" model for face recognition attack post-event screening and classification calculation.

[0074] The dynamic graph attention network model that suppresses oversmoothing is shown in the following formula:

[0075]

[0076] In the formula of the model above, Using dynamic graph attention mechanism for "deep graph neural network" models i，j / a i，i In the "post-recording similarity connected graph data of facial recognition" G simi In this process, data information is transferred and aggregated among all "neighboring nodes" in the direction of "edge".

[0077] Among them, v i With v j Represents any two pairs of "neighbor nodes"; n and m represent G respectively. simi The number of "node" data and "edge" data in the table; Θ s Let Θ be the affine transformation operation matrix. Here, the affine transformation operation matrix Θ needs to undergo an s-th power self-product operation. This self-product operation can effectively alleviate and suppress the "oversmoothing problem" in the convolution operation of traditional graph neural networks. s can be considered as the oversmoothing suppression parameter.

[0078] In addition, v′ i This indicates that G represents the "post-recording similarity connected graph data of facial recognition" simi In this context, each "node" data is the result of data information transfer and aggregation calculation among all "neighboring nodes" in the "edge" direction.

[0079] In the "Dynamic Graph Attention Network Model for Suppressing Oversmoothing," the dynamic graph attention mechanism a i，j The specific calculation process is shown in the following formula:

[0080]

[0081] Attention mechanism in the above animated diagram i，j In the calculation formula, v j Indicates the current "node" v i Any pair of "neighbor node" data, and v k Indicates the current "node" v i All pairs of n-1 "neighbor node" data; E i，j / E i，kThis represents the feature E of the "edge" data used to connect pairs of "neighbor node" data. T Both Θ and Θ represent affine transformation operations in fully connected layers; LeakyReLU is the activation function; and exp represents exponential operations with the natural number e as the base.

[0082] Finally, the "face recognition post-recording similarity connected graph data" G simi In the process, the result v′ of each “node” data is obtained by transferring and aggregating data information among all “neighboring nodes” along the “edge” direction. i The data is input into the screening and classification layer network model to complete the classification calculation for post-attack screening of face recognition attacks. The specific calculation process of the screening and classification layer network model is shown in the following formula:

[0083] r′ i =Θ cis (LeakyReLU(v′ i ));

[0084] In the formula above, LeakyReLU is the activation function; Θ cls This is a fully connected affine transformation network used for the final screening and classification calculation; r′ i This refers to "post-record similarity connected graph data of facial recognition" G. simi In the middle, each "node" data v i The results of post-attack screening and classification calculations for facial recognition attacks.

[0085] This embodiment of a face recognition attack processing method calculates face image similarity values ​​by using a pre-set face similarity calculation model to obtain face metadata and post-event face record data corresponding to the face. Then, based on the face post-event record data and face image similarity values, a face recognition post-event record similarity connected graph data is constructed, with state information record data corresponding to the face as node data and face image similarity values ​​as edge data. Finally, the face recognition post-event record similarity connected graph data is input into a pre-set deep graph neural network model for face recognition attack post-event screening and classification calculation. This solves the technical problem that currently used face recognition attacks involving masks, face stickers, etc., which are not performed on the user's face, often fail to accurately identify many non-human attack cases.

[0086] Furthermore, this application employs a dynamic graph attention network model that suppresses oversmoothing, which can accurately screen, classify, and identify post-event facial recognition records that are either "not the person in question" or "electronic screen photocopying" type black market facial recognition attacks.

[0087] In this embodiment, the interpretability of constructing "post-face recognition record similarity connected graph data" as input into a "deep graph neural network" model (a dynamic graph attention network that suppresses oversmoothing) for post-face recognition attack screening and classification can be summarized as follows: In the "post-face recognition record similarity connected graph data", among the multiple face recognition record "node" data of each customer, most or all of the features of the normal face recognition record "node" data are similar or convergent, because they all come from the face recognition records of the same customer. If a single customer's multiple facial recognition record "nodes" contain records of malicious attacks on the customer's bank account, the characteristics of these malicious attacks will differ significantly from the characteristics of most of the customer's other normal facial recognition record "nodes." The "deep graph neural network" model can accurately screen, classify, and identify which facial recognition record "nodes" are "not the person in question" or "electronic screen re-photographed" malicious facial recognition attack records based on the transmission and aggregation of data features between all pairs of "neighbor nodes" in the "facial recognition post-record similarity connected graph data."

[0088] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a face recognition attack device according to an embodiment of this application, such as... Figure 3 As shown, specifically:

[0089] The similarity calculation unit 301 is used to obtain the instruction to intercept media data in the communication interaction mode;

[0090] Similarity connectivity graph unit 302 is used to generate media files of the corresponding format from the media data of the intercepted region according to the intercepted media data instruction;

[0091] The screening and classification calculation unit 303 is used to cache media files in a preset memory and forward the media files to another interactive terminal in the communication interaction mode through the preset memory.

[0092] In this embodiment, the similarity calculation unit 301 calculates the face image similarity value based on the acquired face metadata and post-face record data corresponding to the face using a preset face similarity calculation model. The similarity connected graph unit 302 then constructs a face recognition post-record similarity connected graph data based on the face post-record data and the face image similarity value. The graph data contains state information record data corresponding to the face as node data and face image similarity value as edge data. Finally, the screening and classification calculation unit 303 inputs the face recognition post-record similarity connected graph data into a preset deep graph neural network model to perform face recognition attack post-screening and classification calculation. This solves the technical problem that currently used face recognition attacks involving non-human entities such as those wearing masks or face stickers fail to accurately identify many non-human attack cases in post-screening and defense.

[0093] Furthermore, this embodiment employs a dynamic graph attention network model that suppresses oversmoothing, which can accurately screen, classify, and identify post-event facial recognition records that are either "not the person in question" or "electronic screen photocopying" type black market facial recognition attacks.

[0094] In this embodiment, the interpretability of constructing "post-face recognition record similarity connected graph data" as input into a "deep graph neural network" model (a dynamic graph attention network that suppresses oversmoothing) for post-face recognition attack screening and classification can be summarized as follows: In the "post-face recognition record similarity connected graph data", among the multiple face recognition record "node" data of each customer, most or all of the features of the normal face recognition record "node" data are similar or convergent, because they all come from the face recognition records of the same customer. If a single customer's multiple facial recognition record "nodes" contain records of malicious attacks on the customer's bank account, the characteristics of these malicious attacks will differ significantly from the characteristics of most of the customer's other normal facial recognition record "nodes." The "deep graph neural network" model can accurately screen, classify, and identify which facial recognition record "nodes" are "not the person in question" or "electronic screen re-photographed" malicious facial recognition attack records based on the transmission and aggregation of data features between all pairs of "neighbor nodes" in the "facial recognition post-record similarity connected graph data."

[0095] This application also provides another face recognition attack device, such as... Figure 4As shown, for ease of explanation, only the parts related to the embodiments of this application are shown. For specific technical details not disclosed, please refer to the method section of the embodiments of this application. The terminal can be any terminal device including mobile phones, tablets, personal digital assistants (PDAs), point-of-sale (POS) terminals, in-vehicle computers, etc. Taking a mobile phone as an example:

[0096] Figure 4 This is a block diagram illustrating a portion of the structure of a mobile phone related to the terminal provided in the embodiments of this application. (Reference) Figure 4 The mobile phone includes: a radio frequency (RF) circuit 1010, a memory 1020, an input unit 1030, a display unit 1040, a sensor 1050, an audio circuit 1060, a wireless fidelity (WiFi) module 1070, a processor 1080, and a power supply 1090, etc. Those skilled in the art will understand that... Figure 4 The mobile phone structure shown does not constitute a limitation on the mobile phone and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0097] The following is combined Figure 4 A detailed introduction to each component of a mobile phone:

[0098] The RF circuit 1010 can be used for receiving and transmitting signals during information transmission or calls. Specifically, it receives downlink information from the base station and processes it with the processor 1080; additionally, it transmits uplink data to the base station. Typically, the RF circuit 1010 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low-noise amplifier (LNA), a duplexer, etc. Furthermore, the RF circuit 1010 can also communicate wirelessly with networks and other devices. The aforementioned wireless communications may use any communication standard or protocol, including but not limited to Global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, and Short Messaging Service (SMS).

[0099] The memory 1020 can be used to store software programs and modules. The processor 1080 executes various mobile phone functions and data processing by running the software programs and modules stored in the memory 1020. The memory 1020 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 1020 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0100] The input unit 1030 can be used to receive input numerical or character information, and to generate key signal inputs related to user settings and function control of the mobile phone. Specifically, the input unit 1030 may include a touch panel 1031 and other input devices 1032. The touch panel 1031, also known as a touch screen, can collect touch operations performed by the user on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel 1031), and drive the corresponding connection devices according to a pre-set program. Optionally, the touch panel 1031 may include two parts: a touch detection device and a touch controller. The touch detection device detects the user's touch position and the signal generated by the touch operation, and transmits the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, and sends it to the processor 1080, and can also receive and execute commands sent by the processor 1080. In addition, the touch panel 1031 can be implemented using various types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch panel 1031, the input unit 1030 may also include other input devices 1032. Specifically, other input devices 1032 may include, but are not limited to, one or more of the following: physical keyboard, function keys (such as volume control buttons, power buttons, etc.), trackball, mouse, joystick, etc.

[0101] The display unit 1040 can be used to display information input by the user or information provided to the user, as well as various menus of the mobile phone. The display unit 1040 may include a display panel 1041, which may optionally be configured as a Liquid Crystal Display (LCD), Organic Light-Emitting Diode (OLED), or similar display panel 1041. Further, a touch panel 1031 may cover the display panel 1041. When the touch panel 1031 detects a touch operation on or near it, it transmits the information to the processor 1080 to determine the type of touch event. Subsequently, the processor 1080 provides corresponding visual output on the display panel 1041 according to the type of touch event. Although in Figure 3 In this embodiment, the touch panel 1031 and the display panel 1041 are two separate components to realize the input and output functions of the mobile phone. However, in some embodiments, the touch panel 1031 and the display panel 1041 can be integrated to realize the input and output functions of the mobile phone.

[0102] The mobile phone may also include at least one sensor 1050, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor. The ambient light sensor can adjust the brightness of the display panel 1041 according to the ambient light level, and the proximity sensor can turn off the display panel 1041 and / or the backlight when the phone is moved to the ear. As a type of motion sensor, an accelerometer sensor can detect the magnitude of acceleration in various directions (generally three axes). When stationary, it can detect the magnitude and direction of gravity and can be used for applications that recognize the phone's posture (such as landscape / portrait switching, related games, magnetometer posture calibration), vibration recognition-related functions (such as pedometer, taps), etc. Other sensors that may be configured in the mobile phone, such as gyroscopes, barometers, hygrometers, thermometers, and infrared sensors, will not be described in detail here.

[0103] The audio circuit 1060, speaker 1061, and microphone 1062 provide an audio interface between the user and the mobile phone. The audio circuit 1060 converts the received audio data into electrical signals and transmits them to the speaker 1061, where the speaker 1061 converts them into sound signals for output. On the other hand, the microphone 1062 converts the collected sound signals into electrical signals, which are then received by the audio circuit 1060, converted into audio data, and then processed by the processor 1080 before being transmitted via the RF circuit 1010 to, for example, another mobile phone, or the audio data can be output to the memory 1020 for further processing.

[0104] WiFi is a short-range wireless transmission technology. Through the WiFi module 1070, mobile phones can help users send and receive emails, browse web pages, and access streaming media, providing users with wireless broadband internet access. Although Figure 3 The WiFi module 1070 is shown, but it is understood that it is not an essential component of a mobile phone and can be omitted as needed without changing the essence of the invention.

[0105] The processor 1080 is the control center of the mobile phone, connecting various parts of the phone through various interfaces and lines. It executes software programs and / or modules stored in the memory 1020 and calls data stored in the memory 1020 to perform various functions and process data, thereby providing overall monitoring of the phone. Optionally, the processor 1080 may include one or more processing units; preferably, the processor 1080 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 1080.

[0106] The mobile phone also includes a power supply 1090 (such as a battery) that supplies power to various components. Preferably, the power supply can be logically connected to the processor 1080 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system.

[0107] Although not shown, mobile phones may also include a camera, Bluetooth module, etc., which will not be described in detail here.

[0108] In this embodiment of the application, the processor 1080 included in the terminal also has the following functions:

[0109] S1. Calculate the face image similarity value by using a pre-set face similarity calculation model to obtain the face metadata and post-face record data corresponding to the face.

[0110] S2. Based on the face post-record data and the face image similarity value, construct a face recognition post-record similarity connected graph data, with state information recording data corresponding to the face as node data and the face image similarity value as edge data;

[0111] S3. Input the face recognition post-record similarity connected graph data into a preset deep graph neural network model to perform face recognition attack post-screening and classification calculation.

[0112] This application also provides a computer-readable storage medium for storing program code that executes any one of the aforementioned implementation methods for handling face recognition attacks.

[0113] This application provides a method, apparatus, device, and storage medium for handling facial recognition attacks. By obtaining a media data interception instruction in a communication interaction mode, and then generating a media file of the corresponding format for the media data of the intercepted area according to the media data interception instruction, the media file is finally cached in a preset memory and forwarded to another interactive terminal in the communication interaction mode through the preset memory. This solves the technical problem that currently, in the post-event screening and defense of facial recognition attacks using masks, facial stickers, etc., which are not performed on the user's own face, many cases of non-human attacks are not accurately identified.

[0114] Furthermore, the interface input parameters required for automated test cases in the technical solution provided in this application embodiment can be automatically adapted and generated, reducing the need for testers to synchronously adjust test case input parameters to adapt to the tested environment due to changes in the environment; the technical solution provided in this application embodiment can reduce the difficulty of data generation for testers, eliminating the need for manual data generation in each environment's dependent third-party system, as the system will automatically read or generate the data generated by the third-party system required for the test cases as input parameters; at the same time, testers do not need to hardcode response result assertions, and can automatically calculate and match according to the rules of different input parameters and interface generated results, thus automatically comparing whether the response message fields and table information are correct.

[0115] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0116] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification 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, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “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.

[0117] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0119] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0120] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0121] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0122] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for handling face recognition attacks, characterized in that, include: S1. Calculate the face image similarity value by using a pre-set face similarity calculation model to obtain the face metadata and post-face record data corresponding to the face. S2. Based on the face post-record data and the face image similarity value, construct a face recognition post-record similarity connected graph data, which contains state information recording data corresponding to the face as node data and face image similarity value as edge data, including both node and edge data. S3. Input the face recognition post-event record similarity connected graph data into a preset deep graph neural network model to perform face recognition attack post-event screening and classification calculation. Step S1 specifically includes: S11. Extract face images from the face metadata and the face post-recording data to obtain at least two face recognition image data; S12. The face recognition image data is paired up in pairs, and face key points are identified according to the landmark algorithm, as well as face pose correction and face feature extraction processing are performed. S13. The facial recognition image data after the facial key point recognition, facial pose correction, and facial feature extraction processing is used to calculate the facial image similarity value through the preset facial similarity calculation model.

2. The face recognition attack processing method according to claim 1, characterized in that, The procedure preceding step S1 also includes: S4. Obtain the face metadata corresponding to the face, and the face post-recording data corresponding to the preset business scenario.

3. The face recognition attack processing method according to claim 2, characterized in that, Step S3 specifically includes: S31. Input the face recognition post-recording similarity connected graph data into a dynamic graph attention network model to suppress oversmoothing. S32. Using the dynamic graph attention network model that suppresses oversmoothing, the similarity connected graph data of the face recognition post-record is processed to perform information transfer and aggregation calculations on the neighbor node data between each pair of neighbors of the node data in the edge direction corresponding to the edge data. S33. Input the data information transmission and aggregation calculation results of the neighbor node data into the screening classification layer network model to perform post-face recognition attack screening and classification calculation.

4. The face recognition attack processing method according to claim 1, characterized in that, The pre-set face similarity calculation model includes: Among them, i o / i1 represents two paired facial recognition image data; CNN n It is a multi-layer convolutional neural network; These are the data features extracted from two paired face recognition image data; similarity dual The facial image similarity value is the final calculated facial image similarity value between two facial recognition image data.

5. The face recognition attack processing method according to claim 3, characterized in that, The dynamic graph attention network model for suppressing oversmoothing includes: Among them, v j For the current node v i The data of any pair of neighboring nodes, and v k For the current node v i All pairs of n-1 neighbor node data, E i，j / E i，k For the feature E of the edge data connecting the paired neighbor node data, Θ represents the affine transformation operation of the fully connected layer network, LeakyReLU is the activation function, and exp represents the exponential operation with the natural number e as the base.

6. The face recognition attack processing method according to any one of claims 1 to 5, characterized in that, The face metadata includes static face recognition associated data; The post-event facial recognition data includes dynamic scene facial recognition data associated with pre-defined business scenarios.

7. A facial recognition attack processing device, characterized in that, include: The similarity calculation unit is used to calculate the face image similarity value based on the obtained face metadata and post-face record data corresponding to the face through a pre-set face similarity calculation model. The similarity connected graph unit is used to construct a face recognition post-record similarity connected graph data based on the face post-record data and the face image similarity value, with the state information record data corresponding to the face as node data and the face image similarity value as edge data; The screening and classification calculation unit is used to input the face recognition post-event record similarity connected graph data into a preset deep graph neural network model to perform face recognition attack post-event screening and classification calculation.

8. A face recognition attack processing device, characterized in that, The device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the face recognition attack processing method according to any one of claims 1-6 according to the instructions in the program code.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program code for executing the face recognition attack processing method according to any one of claims 1-6.

Images