CN114817937A - Keyboard encryption method, device, storage medium and computer program product - Google Patents
Keyboard encryption method, device, storage medium and computer program product Download PDFInfo
- Publication number
- CN114817937A CN114817937A CN202110069624.9A CN202110069624A CN114817937A CN 114817937 A CN114817937 A CN 114817937A CN 202110069624 A CN202110069624 A CN 202110069624A CN 114817937 A CN114817937 A CN 114817937A
- Authority
- CN
- China
- Prior art keywords
- image
- keyboard
- character
- character image
- encrypted
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 79
- 238000004590 computer program Methods 0.000 title claims abstract description 27
- 238000013528 artificial neural network Methods 0.000 claims abstract description 74
- 230000004927 fusion Effects 0.000 claims abstract description 27
- 230000000007 visual effect Effects 0.000 claims abstract description 26
- 238000012549 training Methods 0.000 claims description 69
- 230000008569 process Effects 0.000 claims description 19
- 238000010276 construction Methods 0.000 claims description 4
- 230000006399 behavior Effects 0.000 abstract description 15
- 230000006870 function Effects 0.000 description 35
- 238000013145 classification model Methods 0.000 description 28
- 238000003062 neural network model Methods 0.000 description 20
- 238000010586 diagram Methods 0.000 description 18
- 238000000605 extraction Methods 0.000 description 5
- 230000002452 interceptive effect Effects 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 102100023927 Asparagine synthetase [glutamine-hydrolyzing] Human genes 0.000 description 2
- 101100380329 Homo sapiens ASNS gene Proteins 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 101100048480 Vaccinia virus (strain Western Reserve) UNG gene Proteins 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/602—Providing cryptographic facilities or services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/70—Protecting specific internal or peripheral components, in which the protection of a component leads to protection of the entire computer
- G06F21/82—Protecting input, output or interconnection devices
- G06F21/83—Protecting input, output or interconnection devices input devices, e.g. keyboards, mice or controllers thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2221/00—Indexing scheme relating to security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F2221/21—Indexing scheme relating to G06F21/00 and subgroups addressing additional information or applications relating to security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F2221/2107—File encryption
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computer Hardware Design (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Computer Security & Cryptography (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioethics (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Character Discrimination (AREA)
- Input From Keyboards Or The Like (AREA)
Abstract
The disclosed embodiments relate to a keyboard encryption method, device, storage medium and computer program product. The method comprises the following steps: extracting the characteristics of the keyboard character image, determining fusion characteristics according to the characteristics of the keyboard character image and the characteristics of wrong keyboard characters, and determining an encrypted keyboard character image according to the fusion characteristics; the wrong keyboard character is different from the character corresponding to the keyboard character image; constructing an encrypted keyboard based on the encrypted keyboard character image; the neural network recognition result of the encrypted keyboard character image is an error keyboard character, and the visual recognition results of the encrypted keyboard character image and the keyboard character image are the same. By adopting the method, the safety of the keyboard can be improved, and malicious registration behaviors can be effectively blocked.
Description
Technical Field
The present disclosure relates to the field of communications technologies, and in particular, to a keyboard encryption method, device, storage medium, and computer program product.
Background
With the rapid development of communication electronic technology, smart phones are widely used. The smart phone can be installed with various Applications (APPs) to provide various internet services to the user, and the user usually inputs a user name, a password and an authentication code by using a numeric keypad when registering or logging in the APP.
However, malicious black products can often control a click device to click a numeric keyboard to perform malicious registration, thereby interfering normal operation of enterprises. Even if the smart phone uses the random numeric keyboard, a malicious attacker can also adopt a character recognition algorithm to recognize characters on the numeric keyboard to carry out malicious registration.
Disclosure of Invention
The embodiment of the disclosure provides a keyboard encryption method, a device, a storage medium and a computer program product, which can improve the security of a keyboard and effectively block malicious registration behaviors.
In a first aspect, an embodiment of the present disclosure provides a keyboard encryption method, where the method includes:
extracting the characteristics of the keyboard character image, determining fusion characteristics according to the characteristics of the keyboard character image and the characteristics of the wrong keyboard character, and determining the encrypted keyboard character image according to the fusion characteristics; the wrong keyboard character is different from the character corresponding to the keyboard character image;
constructing an encrypted keyboard based on the encrypted keyboard character image; the neural network recognition result of the encrypted keyboard character image is an error keyboard character, and the visual recognition results of the encrypted keyboard character image and the keyboard character image are the same.
In a second aspect, an embodiment of the present disclosure provides a computer device, including:
the encryption unit is used for extracting the characteristics of the keyboard character image, determining fusion characteristics according to the characteristics of the keyboard character image and the characteristics of wrong keyboard characters, and determining the encrypted keyboard character image according to the fusion characteristics; the wrong keyboard character is different from the character corresponding to the keyboard character image;
the construction unit is used for constructing an encrypted keyboard based on the encrypted keyboard character image; the neural network recognition result of the encrypted keyboard character image is an error keyboard character, and the visual recognition results of the encrypted keyboard character image and the keyboard character image are the same.
In a third aspect, an embodiment of the present disclosure provides a server, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the method of the first aspect when executing the computer program.
In a fourth aspect, the embodiments of the present disclosure provide a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the method of the first aspect described above.
In a fifth aspect, the embodiments of the present disclosure provide a computer program product comprising a computer program which, when executed by a processor, implements the method of the first aspect described above.
The keyboard encryption method, the computer device and the storage medium provided by the embodiment of the disclosure can add the characteristics of the error characters (for example, the error keyboard characters) in the original character image (for example, the keyboard character image) to interfere the identification result of the attack party neural network model, help the server to identify the machine registration behavior, effectively block the malicious registration behavior and improve the security of the keyboard. Meanwhile, the loss value of the encrypted character image and the loss value of the original character image are smaller than a preset value, so that the encrypted character image and the original character image can be ensured to have the same visual identification result, namely the encrypted character image still looks like the original character image by human eyes, and the human eye identification of a user cannot be influenced.
Drawings
Fig. 1 is an application environment diagram of a keyboard encryption method provided in the embodiment of the present disclosure;
FIG. 2 is a schematic diagram of a keyboard application provided by an embodiment of the present disclosure;
fig. 3 is a schematic flowchart of a keyboard encryption method according to an embodiment of the present disclosure;
FIG. 4 is a schematic diagram of a character image provided by an embodiment of the present disclosure;
fig. 5 is a schematic diagram of transmission of keyboard image information according to an embodiment of the present disclosure;
FIG. 6 is a schematic diagram of another transmission of keyboard image information according to an embodiment of the present disclosure;
fig. 7 is another schematic flow chart of a keyboard encryption method according to an embodiment of the present disclosure;
FIG. 8 is a schematic diagram of an image encryption model provided by an embodiment of the present disclosure;
fig. 9 is a schematic diagram of an encryption effect provided by an embodiment of the present disclosure;
FIG. 10 is a schematic diagram of a model training process provided by an embodiment of the present disclosure;
FIG. 11 is a schematic diagram of a neural network model provided by an embodiment of the present disclosure;
FIG. 12 is a schematic diagram of a loss function provided by an embodiment of the present disclosure;
FIG. 13 is a schematic diagram of model training provided by embodiments of the present disclosure;
FIG. 14 is another schematic diagram of a model training process provided by embodiments of the present disclosure;
FIG. 15 is a schematic diagram of a training data set provided by an embodiment of the present disclosure;
FIG. 16 is a schematic diagram of the encryption effect provided by the embodiment of the disclosure;
FIG. 17 is a block diagram of a computer device provided by an embodiment of the present disclosure;
FIG. 18 is another block diagram of a computer device provided by an embodiment of the present disclosure;
fig. 19 is an internal structural diagram of a computer device provided in the embodiment of the present disclosure.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present disclosure more clearly understood, the embodiments of the present disclosure are described in further detail below with reference to the accompanying drawings and the embodiments. It is to be understood that the specific embodiments described herein are merely illustrative of the embodiments of the disclosure and that no limitation to the embodiments of the disclosure is intended.
First, before specifically describing the technical solution of the embodiment of the present disclosure, a technical background or a technical evolution context on which the embodiment of the present disclosure is based is described. Typically, a user usually needs to input a user name, a password and an authentication code when registering or logging in the APP to help the server identify machine behaviors, so as to block malicious registered machine behaviors. Based on this background, applicants have found that even using a random keyboard to randomly shuffle the positions of characters, malicious registration behavior is not resisted. The attacker can still identify the character corresponding to each position by using a character identification algorithm to carry out malicious registration. How to resist the malicious registration behavior becomes a difficult problem to be solved urgently at present. In order to solve the problem, the applicant pays a great deal of creative work and proposes the technical scheme introduced by the following embodiment.
The following describes technical solutions related to the embodiments of the present disclosure with reference to a scenario in which the embodiments of the present disclosure are applied.
The keyboard encryption method provided by the embodiment of the disclosure can be applied to the application environment shown in fig. 1. Wherein the electronic device 10 communicates with the server 20 over a network. The electronic device can install the APP to provide various internet services for the user, for example, a network appointment service, a network shopping service and the like are provided. The server 20 may be an application server of the APP for supporting a background implementation of the services provided by the APP.
In specific implementations, the electronic device 10 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and the server 20 may be implemented by an independent server or a server cluster composed of a plurality of servers.
Referring to fig. 2, when a user registers at an APP or web page, a user name, a password, and an authentication code may be input through a keypad (e.g., a numeric keypad) of electronic device 10. The electronic device 10 may send the user name, password, and authentication code entered by the user to the server 20, and the server 20 may complete the user registration based on the received user name, password, and authentication code.
At present, an attacker can control a shocker to click a data keyboard, and sequentially input a user name, a password and an authentication code to carry out malicious registration. In order to improve the security of the keyboard and effectively resist malicious registration behaviors, the embodiment of the disclosure provides a keyboard encryption method. The method is suitable for the system shown in fig. 1, and the execution subject can be the server 20 in the system shown in fig. 1. As shown in fig. 3, the method comprises the steps of:
The keyboard character image is a character image included in the virtual keyboard, and may be a character image of a numeric keyboard, for example. In order to effectively resist malicious registration behaviors, the embodiment of the disclosure aims to add interference features into the features of an original character image and generate an encrypted character image according to the features added with the interference. When encrypting the character image, the server may first perform feature extraction on the character image so as to fuse with the interference features.
For example, each keyboard character image included in the keyboard may be encrypted to improve the security of the keyboard. When encrypting each keyboard character image of the keyboard, firstly, feature extraction is carried out on the keyboard character image, and the features of the keyboard character image are determined.
The keyboard character image corresponds to a character, and the character may be an arabic number, an english alphabet, or a punctuation mark. Referring to fig. 4, "character image 1" corresponds to the character "1", and "character image 2" corresponds to the character "2". It should be noted that the character corresponding to the character image may be referred to as a label of the character image.
In one possible implementation, the features of the keyboard character image may be obtained using an Auto Encoder (AE). The AE includes an encoder (encoder) and a decoder (decoder), wherein the input of the encoder is an image and the output is a feature of the image; the input to the decoder may be a characteristic of the encoder output, the output being an image reconstructed from the input characteristic. In step 201, the server may perform feature extraction on the keyboard character image by using the encoder to obtain features of the keyboard character image. Specifically, a keyboard character image is input to an encoder, the output of which is a characteristic of the keyboard character image.
Wherein the incorrect keyboard character is different from the character actually corresponding to the keyboard character image.
When the server adds interference to the original keyboard character image, an error keyboard character can be designated so as to point the identification result of the neural network model of the attacker to the error keyboard character. Illustratively, when encrypting the keyboard character image, an incorrect keyboard character is specified that is different from the keyboard character image. Fusion characteristics can be determined according to the characteristics of the keyboard character image and the characteristics of the wrong keyboard character, namely the characteristics of the keyboard character image and the characteristics of the second character are fused.
For example, the server may process the features of the keyboard character image and the features of the incorrect keyboard character using a feature fusion algorithm to obtain the above-mentioned fusion features. The feature fusion algorithm includes, but is not limited to, addition, multiplication, and the like of features, which is not limited by the embodiments of the present disclosure.
According to the embodiment of the disclosure, the malicious registration behavior of the attacker is effectively resisted by interfering the recognition result of the neural network model of the attacker. The encrypted keyboard character image is not visually different from the original keyboard character image, that is, the encrypted keyboard character image and the original keyboard character image are not different from each other by human eyes. But the neural network is used for identifying the encrypted keyboard character image, and the identification result is different from the original character.
Illustratively, the neural network recognition result of the encrypted keyboard character image is the second character, but the visual recognition result is still the first character.
In a possible implementation manner, in order to ensure that the visual recognition result of the encrypted keyboard character image is still the character really corresponding to the image, the difference between the encrypted keyboard character image and the original keyboard character image is less than or equal to a preset threshold value, where the preset threshold value is a minimum loss value that ensures that the visual recognition results of the original keyboard character image and the encrypted keyboard character image are the same. The preset threshold may be a threshold set according to experience, or a threshold obtained by iteratively training a neural network model.
After the server encrypts each keyboard character image, the server can also construct a keyboard based on the encrypted keyboard character images. The keyboard includes a plurality of different keyboard character images. Illustratively, referring to fig. 5, after the server constructs the keyboard, image information of the keyboard may be further transmitted to the electronic device. The electronic device may display the encrypted keyboard after decoding the image information of the keyboard. Referring to fig. 5, the data keyboard includes different keyboard character images, e.g., a keyboard character image corresponding to the character "1", a keyboard character image corresponding to the character "2", etc. The keyboard character images included in the data keyboard are all encrypted character images, the encrypted character images contain the characteristics of wrong characters, the identification result of the attacking party neural network model can be interfered sufficiently, the visual identification result is still the real corresponding characters, and the influence on the human eye identification of a user due to the introduction of the interference of wrong keyboard characters can be avoided.
For example, referring to fig. 6, after encrypting each keyboard character image, the server transmits image information of the character image to the electronic device. The electronic device may determine a single character image from image information of the character image, combine a plurality of character images in a specific combination manner to construct an encryption keyboard, or randomly combine a plurality of encrypted character images to construct a keyboard to construct an encryption keyboard.
In the method shown in fig. 3, the server may add the features of the wrong keyboard character to the original keyboard character image to interfere with the recognition result of the aggressor neural network model, help the server to recognize the machine registration behavior, effectively block the malicious registration behavior, and improve the security of the keyboard. Meanwhile, the encrypted keyboard character image has the same visual identification result as the original keyboard character image, namely the encrypted character image still looks like the original character image by human eyes, and the human eye identification of a user cannot be influenced.
In the method provided by the embodiment of the present disclosure, the server may encrypt the original keyboard character image by using an encryption model. Referring to fig. 7, the method specifically includes the following steps:
When encrypting the keyboard image, an incorrect keyboard character may be specified. The features of the keyboard character image are first extracted by an image encoder, and the features of the wrong keyboard character can also be extracted by a character encoder.
In a specific implementation, the server may further train an image encryption model, where the image encryption model includes an image encoder, an image decoder, and a character encoder.
The image encryption model is used for adding the characteristics of error characters in an original character image so as to interfere the recognition result of the neural network model on the character image. For example, the keyboard character image may be encrypted, interfering with the recognition results of the neural network model used by the black product.
Illustratively, referring to fig. 8, the image encoder is used to extract features of a character image, which is input as a character image and output may be a character; the character encoder is used for extracting the character features, the input of the character encoder is the character, and the output of the character encoder is the character features; the image decoder is used for generating an encrypted character image, the input of the encrypted character image is the feature of the original character image fused with the feature of the error character, and the output is the encrypted character image.
In a possible implementation mode, the visual recognition results of an original character image and an encrypted character image are the same as a model training target, and an image encoder and an image decoder are trained; training a character encoder by taking a neural network recognition result of the encrypted character image as an error character as a model training target; the erroneous character is different from the character corresponding to the original character image.
And step 702, fusing the characteristics of the keyboard character image and the characteristics of the wrong keyboard character and inputting the fused keyboard character image into an image decoder of an image encryption model to obtain an encrypted keyboard character image.
For example, referring to fig. 9, the "original keyboard character image 4" is processed by the image encoder to output the feature of the "original keyboard character image 4", the feature is fused with the feature of the error keyboard character "6" and then input to the image decoder, and the encrypted keyboard character image is output. The encrypted keyboard character image still looks like the character '4' to human eyes, but the recognition result is '6' by adopting a deep learning method to recognize the character image, so that the automatic recognition algorithm of an attacker is wrongly recognized, and the encryption effect is achieved.
In the method shown in fig. 7, a special neural network model is trained to generate an encrypted character image, the encrypted character image output by the neural network model can interfere with the recognition result of the neural network model used by an attacker, and meanwhile, the recognition of the user by naked eyes is not affected, and the encrypted character image is visually the same as the original character image.
In the method provided by the embodiment of the present disclosure, the visual recognition results of the original character image and the encrypted character image are the same as the model training target, and the flow of training the image encoder and the image decoder is as shown in fig. 10. Referring to fig. 10, the method specifically includes the following steps:
The self-coding learning module comprises an encoding module (encoder) and a decoding module (decoder) and is used for reconstructing images. For example, the image may be input to the encoding learning module, the image may be encoded by the encoding module to obtain the features of the image, or the features of the image may be input to the decoding module to decode the input features to reconstruct the image. The training target of the self-coding learning module is that the loss value of the input image and the reconstructed image is small enough, and the input image and the reconstructed image are not different from each other visually.
For example, referring to fig. 11, the encoding module includes a convolutional layer and a pooling layer, and the decoding module includes an upsampling layer and a convolutional layer. The input image is subjected to a layer convolution and a pooling layer to obtain the features of the image (which may be referred to as intermediate features). The characteristics of the image pass through the upsampling layer and the convolution layer, and a reconstructed image is output. The Loss function of the reconstructed image and the input image is denoted as Loss 1. And updating the parameters of the self-coding network by using a Loss function Loss 1. After training is finished, an input image is given, and a reconstructed image with the content consistent with that of the original image can be output through the self-coding network.
In a possible implementation, the loss function is used to represent a functional relationship between the loss value of the input image compared to the reconstructed image and the neural network parameters of the encoding module, and the goal of training the self-encoding module is to minimize the loss function. For example, referring to fig. 12, the Loss function Loss 1 may characterize a functional relationship between the coding module neural network and the Loss value, and when the coding module neural network parameter is x, the value of the Loss function Loss 1 is the smallest. The neural network parameter x may represent a set of a series of parameters, or may be a certain parameter, which is not limited in this embodiment of the application. It should be noted that, in the embodiment of the present disclosure, the neural network parameters include, but are not limited to, weights, bias, gradient values, and the like of the neural network.
The encoding module is used for extracting the characteristics of the character image and meeting the requirements of an encryption model, so that the encoding module is determined to be the image encoder.
After the self-encoding module is trained, the neural network parameters of the decoding module can be adjusted, and the decoding module after the parameters are adjusted is the image decoding model in the embodiment of the disclosure. The input of the image decoding model is the feature of the original character image fused with the feature of the error character, and the output of the image decoding model is the encrypted character image of the original character image.
The neural network parameters of the decoding module are adjusted to ensure that the loss values of the encrypted character image and the original character image are small enough and the original character image and the encrypted character image are not visually different from each other.
In one possible implementation, the adjusting of the neural network parameters of the decoding module includes the following steps:
and S1, fusing the characteristics of the original character image and the characteristics of the error character and inputting the fused characters into a decoding module to obtain a reconstructed image.
Specifically, the fused vector of the features of the original character image and the features of the error character is input to the decoding module, and the decoding module can decode an image according to the input features, that is, the reconstructed image.
And S2, determining a first loss function, wherein the first loss function is used for representing the functional relation between the loss value of the original character image compared with the reconstructed image and the neural network parameters of the initial image decoding model.
Wherein, the characters corresponding to the original character image are different from the error characters.
And S3, adjusting the neural network parameters of the image decoding module until the value of the first loss function is the preset threshold value, and determining that the image decoding module is an image decoding model.
For example, referring to fig. 13, the feature of the original character image and the feature of the error character after fusion is v, and the feature is v, and the input image decoding module obtains a reconstructed image. The loss value of the original character image compared with the reconstructed image is loss (x), and when the neural network parameters of the image decoding module change, the loss (x) also changes. When loss (x) takes the minimum value, i.e. the preset threshold value, it is determined that the current image decoding module is the image decoding module, and is used for outputting the encrypted character image.
In the embodiment of the disclosure, the server may train the character encoder to ensure that the features fused with the features of the original character image cause interference to the identification of the aggressor neural network model. Referring to fig. 14, the method specifically includes the following steps:
Wherein the character encoding model includes an embedding (embedding) layer for encoding characters as features.
In one possible implementation, an initial character encoding module may be trained using a deep learning approach. Referring to fig. 15, the training data set includes a large number of character images and labels, where the labels may be considered characters. In the training process, the character image is used as the input of the neural network model, and the cross entropy loss between the output result and the real label of the character image is calculated. And adjusting parameters of the neural network model by taking the minimized cross entropy loss as an optimization target to obtain a stable character coding module capable of accurately identifying characters corresponding to the character image.
And 1402, inputting the error characters into a character coding module to obtain the characteristics of the error characters, fusing the characteristics of the error characters and the characteristics of the original character images, and inputting the fused error characters into the image decoder to obtain a reconstructed image.
In the embodiment of the disclosure, in order to ensure that the neural network model can recognize the encrypted character image as an error character, the image can be reconstructed according to the features fused with the error character to perform iterative training, so that the neural network result of the reconstructed image is continuously close to the error character.
And 1403, adjusting the neural network parameters of the character coding module according to the difference between the neural network identification result of the reconstructed image and the error character until the neural network identification result of the reconstructed image is the error character.
In the concrete implementation, the reconstructed image is input into a character image classification model, a second loss function is determined according to the output of the character image classification model, and the neural network parameters of the initial character coding model are adjusted until the second loss function obtains the minimum value, so that the stable character coder is obtained.
The character image classification model is used for identifying characters corresponding to the character images, the characters are input into the character images, and the characters are output. The reconstructed image is an image obtained by fusing the features of the original character image and the features of the error characters and inputting the fused image into the image decoding model. The second loss function is used for representing the functional relation between the loss value of the output result of the character image classification model compared with the error character and the neural network parameters of the character coding model. The neural network parameters of the character encoding model may be embedding layer parameters.
It should be noted that, when the neural network of the character encoding model changes, the value of the second loss function also changes, that is, the loss value between the output result of the character image classification model and the error character also changes. According to the method, the loss value between the output result of the character image classification model and the error character is reduced through the neural network parameter of the character encoding model, so that the recognition result of the character image classification model on the encrypted character image approaches to the error character, and the recognition result of the character image classification model used by the attacker is effectively interfered.
When the second loss function obtains the minimum value, the recognition result of the character image classification model on the encrypted character image has the highest proximity degree with the error character, and a final character coding model can be determined according to the current neural network parameters and used for coding the error character to obtain the characteristics of the error character.
Illustratively, the second character is input into the character encoding model to obtain the characteristics of the second character. The method can ensure that the character image classification model used by the attacker is effectively interfered after the characteristics of the second character and the characteristics of the keyboard character image are fused.
It should be noted that the character features output by the adjusted character encoding module are beneficial to indicating the visual recognition result of the reconstructed image as an error character, and the requirement of the image encryption model is met.
In the method shown in fig. 14, in order to disturb the recognition result of the neural network model used by the attacker, the features of the wrong character are fused among the features of the original character image. In addition, the character coding model can be trained, and the neural network parameters of the character coding model are adjusted to ensure that the recognition result of the neural network model used by an attacker is different from the real character corresponding to the character image.
The following describes a keyboard encryption method provided by the embodiments of the present disclosure with reference to specific examples. The method specifically comprises a model training process and an application process. Wherein, the training process comprises the following steps:
step a, 'original character image 4' is input to the encoder to obtain the character image feature v 1.
And b, determining an error character '6', and inputting the character '6' into the character coding model. The feature v2 of the character "6" is obtained.
And c, fusing the characteristic v1 with the characteristic v2 to obtain a characteristic v 3.
And d, inputting the characteristic v3 into a decoder to obtain a reconstructed image.
And e, reconstructing an image input character image classification model to obtain a predicted character.
And f, carrying out iterative training to update the imbedding layer parameters of the character coding model and the neural network parameters of the decoder to obtain the stable character coding model and the stable decoder.
Illustratively, a loss function of the reconstructed image and the original character image 4 is calculated, and the neural network parameters of the decoder are adjusted by taking the minimum value of the loss function as an optimization target, so that the loss value of the reconstructed image and the loss value of the original character image 4 are minimum, namely, the reconstructed image and the original character image 4 are the character images corresponding to the character 4 which looks the same to human eyes.
Calculating a loss function of cross entropy loss by calculating the predicted character and the appointed wrong character '6', taking the minimum value of the loss function as an optimization target, and adjusting embedding layer parameters of the character coding model to enable the recognition result of the character image classification model to approach the wrong character '6' as much as possible, thereby achieving the purpose of interfering the recognition result of the neural network model of the attacker.
The application process comprises the following steps: inputting the original character image into an encoder to obtain the characteristics of the original character image, and inputting the error character into a character coding model to obtain the characteristics of the error character. And fusing the characteristics of the original character image and the characteristics of the error character, inputting the fused characters into an image decoder, and outputting an encrypted character image.
For example, referring to fig. 16, assuming that the verification code set by the server at the time of APP registration is 480, the electronic device displays a keyboard constructed by the encrypted character image of the server, and the user can accurately recognize the characters "4", 8 "and" 0 ", and accurately input the verification code" 480 ". However, if the current registration behavior is that of a black product, the neural network model (e.g., a character image classification model) used by the black product cannot recognize the real characters "4", 8 "and" 0 "because the encrypted character image adds the features of an error character. For example, the character "4" is recognized as the character "6" by the black product, the character "6" is erroneously recognized as the character "4", the character "3" is recognized as the character "8", the character "9" is recognized as the character "0", the verification code input by the black product control defibrillator is "639", successful registration cannot be performed, and thus the disclosure can effectively block malicious registration behaviors.
It should be understood that, although the steps in the flowcharts of fig. 3, 7, 10 and 14 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 3, 7, 10, and 14 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least some of the other steps.
The embodiment of the present disclosure provides a computer device, which may be the server 20 according to the embodiment of the present disclosure. As shown in fig. 17, the computer apparatus includes: an encryption unit 1701 and a construction unit 1702.
The encryption unit 1701 is configured to perform feature extraction on the keyboard character image, determine features of the keyboard character image, determine fusion features according to the features of the keyboard character image and features of wrong keyboard characters, and determine an encrypted keyboard character image according to the fusion features; the wrong keyboard character is different from the character corresponding to the keyboard character image;
the constructing unit 1702 is configured to construct an encrypted keyboard based on the encrypted keyboard character image; the neural network recognition result of the encrypted keyboard character image is an error keyboard character, and the visual recognition results of the encrypted keyboard character image and the keyboard character image are the same.
The encryption unit 1701 is specifically configured to input a keyboard character image into an image encoder of the image encryption model to obtain a feature of the keyboard character image, and input an incorrect keyboard character into a character encoder of the image encryption model to obtain a feature of the incorrect keyboard character; and fusing the characteristics of the keyboard character image and the characteristics of the wrong keyboard character, and inputting the fused keyboard character image into an image decoder of an image encryption model to obtain an encrypted keyboard character image.
In one possible implementation, referring to fig. 18, the computer device further comprises a training unit 1703. The training unit 1703 is used for training an image encryption model, which includes an image encoder, an image decoder, and a character encoder.
Illustratively, the training process of the image encryption model includes: training an image encoder and an image decoder by taking the same visual recognition results of the original character image and the encrypted character image as a model training target; training a character encoder by taking a neural network recognition result of the encrypted character image as an error character as a model training target; the erroneous character is different from the character corresponding to the original character image.
In a possible implementation, the training unit 1703 is configured to train an image self-encoder; the image self-encoder comprises an encoding module and a decoding module; the encoding module is used for determining the characteristics of a character image input to the encoder; the decoding module is used for reconstructing a character image according to the characteristics input into the decoding module;
determining an encoding module as an image encoder;
adjusting the neural network parameters of the decoding module to obtain an image decoder; the image decoder is used for outputting the encrypted character image of the original character image, the difference between the encrypted character image of the original character image and the original character image is smaller than or equal to a preset threshold value, and the preset threshold value is used for ensuring that the visual recognition results of the encrypted character image and the original character image are the same.
In a possible implementation manner, the training unit 1703 is specifically configured to fuse the features of the original character image and the features of the error character and input the fused features to a decoder to obtain a reconstructed image;
adjusting the neural network parameters of the decoding module until the first loss function obtains the minimum value, and determining the adjusted decoding module as an image decoder; the first loss function is used for representing the functional relation between the loss value of the original character image compared with the reconstructed image and the neural network parameters of the decoding model.
In a possible implementation manner, the training unit 1703 is further configured to train the character encoding module, and input the error character into the character encoding module to obtain a feature of the error character; the character coding module is used for extracting the characteristics of the characters;
fusing the characteristics of the error characters and the characteristics of the original character images and inputting the fused characters into an image decoder to obtain a reconstructed image;
adjusting the neural network parameters of the character coding module according to the difference between the neural network recognition result of the reconstructed image and the error character until the neural network recognition result of the reconstructed image is the error character;
and determining the adjusted character encoding module as a character encoder.
In a possible implementation manner, the training unit 1703 is specifically configured to input the reconstructed image into a character image classification model, and determine a second loss function according to an output of the character image classification model; the second loss function is used for representing and determining the functional relation between the loss value of the output result of the character image classification model compared with the error character and the neural network parameter of the character encoder, the input of the character image classification model is a character image, and the output of the character image classification model is a character; and adjusting the neural network parameters of the character encoder until the second loss function obtains a minimum value.
It should be noted that the training of the image encryption model may be implemented by the computer device, for example, by the training unit 1703, or the other device may train the image encryption model and send the trained image encryption model to the computer device.
Fig. 19 is a block diagram illustrating a server 1900 in accordance with an example embodiment. Referring to fig. 19, server 1900 includes processing components 1920, which further include one or more processors, and memory resources, represented by memory 1922, for storing instructions or computer programs, such as application programs, that are executable by processing components 1920. The application programs stored in memory 1922 may include one or more modules each corresponding to a set of instructions. Further, the processing component 1920 is configured to execute instructions to perform the keyboard encryption method described above.
The server 1900 may also include a power component 1924 configured to perform power management of the device 1900, a wired or wireless network interface 1926 configured to connect the device 1900 to a network, and an input/output (I/O) interface 1928. The server 1900 may operate based on an operating system stored in memory 1922, such as Window 1919 over, Mac O19 XTM, UnixTM, LinuxTM, FreeB19DTM, or the like.
In an exemplary embodiment, the present disclosure also provides a computer program product, which when executed by a processor, may implement the above-described method. The computer program product includes one or more computer instructions. When loaded and executed on a computer, may implement some or all of the above-described methods, in whole or in part, according to the procedures or functions described in the embodiments of the disclosure.
Embodiments of the present disclosure also provide a non-transitory computer-readable storage medium, such as a memory, comprising instructions executable by a processor of a computer device (e.g., a server as described above) to perform the above-described method. For example, the non-transitory computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
Embodiments of the present disclosure also provide a storage medium comprising instructions, such as a memory comprising instructions, executable by a processor of a server to perform the above-described method. The storage medium may be a non-transitory computer readable storage medium, which may be, for example, a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
The embodiment of the application discloses a TS1 and a keyboard encryption method, which is characterized by comprising the following steps:
extracting features of the keyboard character image, determining the features of the keyboard character image, determining fusion features according to the features of the keyboard character image and the features of wrong keyboard characters, and determining an encrypted keyboard character image according to the fusion features; the wrong keyboard character is different from the character corresponding to the keyboard character image;
constructing an encrypted keyboard based on the encrypted keyboard character image; and the neural network recognition result of the encrypted keyboard character image is the wrong keyboard character, and the visual recognition results of the encrypted keyboard character image and the keyboard character image are the same.
TS2, the method of TS1, wherein the extracting features of keyboard character images, determining features of the keyboard character images, determining fusion features according to the features of the keyboard character images and features of incorrect keyboard characters, determining encrypted keyboard character images according to the fusion features, comprises:
inputting the keyboard character image into an image encoder of an image encryption model to obtain the characteristics of the keyboard character image, and inputting the wrong keyboard character into a character encoder of the image encryption model to obtain the characteristics of the wrong keyboard character;
and fusing the characteristics of the keyboard character image and the characteristics of the wrong keyboard character and inputting the fused keyboard character image and the fused keyboard character image into an image decoder of the image encryption model to obtain the encrypted keyboard character image.
training the character encoder by taking the neural network recognition result of the encrypted character image as an error character as a model training target; the error character is different from the character corresponding to the original character image.
The TS 4, as set forth in TS 3, is a method for training the image encoder and the image decoder by using a model training target in which visual recognition results of an original character image and an encrypted character image are the same, and includes: training an image self-encoder; the image self-encoder comprises an encoding module and a decoding module; the encoding module is used for determining the characteristics of a character image input to the encoder; the decoding module is used for reconstructing a character image according to the characteristics input into the decoding module;
determining that the encoding module is the image encoder;
adjusting neural network parameters of the decoding module to obtain the image decoder; the image decoder is used for outputting an encrypted character image of the original character image, the difference between the encrypted character image of the original character image and the original character image is smaller than or equal to a preset threshold value, and the preset threshold value is used for ensuring that the visual recognition results of the encrypted character image and the original character image are the same.
The method of TS 5, as set forth in TS 4, wherein the adjusting the neural network parameters of the decoding module to obtain the image decoder comprises: fusing the characteristics of the original character image and the characteristics of the error character and inputting the fused characters into the decoder to obtain a reconstructed image;
adjusting the neural network parameters of the decoding module until the first loss function obtains the minimum value, and determining the adjusted decoding module as the image decoder; the first loss function is used for representing the functional relation between the loss value of the original character image compared with the reconstructed image and the neural network parameters of the decoding model.
The method of TS 6 and TS 3, wherein the training of the character encoder with the neural network recognition result of the encrypted character image as an error character as a model training target includes: the training character coding module inputs the error characters into the character coding module to obtain the characteristics of the error characters; the character coding module is used for extracting the characteristics of the characters;
fusing the characteristics of the error characters and the characteristics of the original character images and inputting the fused characters into the image decoder to obtain a reconstructed image;
adjusting the neural network parameters of the character coding module according to the difference between the neural network recognition result of the reconstructed image and the error character until the neural network recognition result of the reconstructed image is the error character;
and determining the adjusted character encoding module as the character encoder.
The method of TS7, as set forth in TS 6, wherein the adjusting the neural network parameters of the character encoding module according to the difference between the neural network identification result of the reconstructed image and the error character until the neural network identification result of the reconstructed image is the error character includes: inputting the reconstructed image into a character image classification model, and determining a second loss function according to the output of the character image classification model; the second loss function is used for representing a functional relation between the loss value of the output result of the character image classification model compared with the error character and the neural network parameter of the character encoder, the input of the character image classification model is a character image, and the output of the character image classification model is a character;
and adjusting the neural network parameters of the character encoder until the second loss function obtains a minimum value.
TS8, a computer device, characterized in that the device comprises:
the encryption unit is used for extracting the characteristics of the keyboard character image, determining fusion characteristics according to the characteristics of the keyboard character image and the characteristics of wrong keyboard characters, and determining the encrypted keyboard character image according to the fusion characteristics; the wrong keyboard character is different from the character corresponding to the keyboard character image;
the construction unit is used for constructing an encrypted keyboard based on the encrypted keyboard character image; and the neural network recognition result of the encrypted keyboard character image is the wrong keyboard character, and the visual recognition results of the encrypted keyboard character image and the keyboard character image are the same.
TS9, the device of TS8, wherein the encryption unit is specifically configured to input the keyboard character image into the image encoder to obtain the characteristics of the keyboard character image, and to input the incorrect keyboard character into the character encoder to obtain the characteristics of the incorrect keyboard character; and fusing the characteristics of the keyboard character image and the characteristics of the wrong keyboard character and inputting the fused keyboard character image and the fused keyboard character image into the image decoder to obtain the encrypted keyboard character image.
TS10, the device of TS9, wherein the training process of the image encryption model comprises: training the image encoder and the image decoder by taking the same visual recognition results of the original character image and the encrypted character image as a model training target;
training the character encoder by taking the neural network recognition result of the encrypted character image as an error character as a model training target; the error character is different from the character corresponding to the original character image.
TS11, the apparatus according to TS10, wherein the training process for training the image encoder and the image decoder specifically includes: training an image self-encoder; the image self-encoder comprises an encoding module and a decoding module; the encoding module is used for determining the characteristics of a character image input to the encoder; the decoding module is used for reconstructing a character image according to the characteristics input into the decoding module;
determining that the encoding module is the image encoder;
adjusting neural network parameters of the decoding module to obtain the image decoder; the image decoder is used for outputting an encrypted character image of the original character image, the difference between the encrypted character image of the original character image and the original character image is smaller than or equal to a preset threshold value, and the preset threshold value is used for ensuring that the visual recognition results of the encrypted character image and the original character image are the same.
TS12, the device according to TS11, wherein the training process of adjusting the neural network parameters of the decoding module to obtain the image decoder specifically comprises: fusing the characteristics of the original character image and the characteristics of the error character and inputting the fused characters into the decoder to obtain a reconstructed image;
adjusting the neural network parameters of the decoding module until the first loss function obtains the minimum value, and determining the adjusted decoding module as the image decoder; the first loss function is used for representing the functional relation between the loss value of the original character image compared with the reconstructed image and the neural network parameters of the decoding model.
TS13 the apparatus of TS10, wherein the training process of training the character encoder specifically includes, with the neural network recognition result of the encrypted character image as an error character as a model training target: the training character coding module inputs the error characters into the character coding module to obtain the characteristics of the error characters; the character coding module is used for extracting the characteristics of the characters;
fusing the characteristics of the error characters and the characteristics of the original character images and inputting the fused characters into the image decoder to obtain a reconstructed image;
adjusting the neural network parameters of the character coding module according to the difference between the neural network recognition result of the reconstructed image and the error character until the neural network recognition result of the reconstructed image is the error character;
and determining the adjusted character encoding module as the character encoder.
TS14, the device according to TS13, wherein the training process of adjusting the neural network parameters of the decoding module to obtain the image decoder specifically comprises: inputting the reconstructed image into a character image classification model, and determining a second loss function according to the output of the character image classification model; the second loss function is used for representing a functional relation between the loss value of the output result of the character image classification model compared with the error character and the neural network parameter of the character encoder, the input of the character image classification model is a character image, and the output of the character image classification model is a character;
and adjusting the neural network parameters of the character encoder until the second loss function obtains a minimum value.
TS15, a server comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method described in TS1 to TS7 when executing the computer program.
TS16, a computer readable storage medium having a computer program stored thereon, the computer program, when being executed by a processor, performing the steps of the method of TS1 to TS 7.
TS17, a computer program product comprising a computer program, characterized in that the computer program realizes the steps of the method described above with reference to TS1 to TS7 when being executed by a processor.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided by the embodiments of the disclosure may include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express a few implementation modes of the embodiments of the present disclosure, and the description thereof is specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for those skilled in the art, variations and modifications can be made without departing from the concept of the embodiments of the present disclosure, and these are all within the scope of the embodiments of the present disclosure. Therefore, the protection scope of the patent of the embodiment of the disclosure should be subject to the appended claims.
Claims (10)
1. A keyboard encryption method, the method comprising:
extracting features of the keyboard character image, determining the features of the keyboard character image, determining fusion features according to the features of the keyboard character image and the features of wrong keyboard characters, and determining an encrypted keyboard character image according to the fusion features; the wrong keyboard character is different from the character corresponding to the keyboard character image;
constructing an encrypted keyboard based on the encrypted keyboard character image; and the neural network recognition result of the encrypted keyboard character image is the wrong keyboard character, and the visual recognition results of the encrypted keyboard character image and the keyboard character image are the same.
2. The method of claim 1, wherein the extracting features of the keyboard character image, determining fusion features according to the features of the keyboard character image and features of incorrect keyboard characters, and determining an encrypted keyboard character image according to the fusion features comprises:
inputting the keyboard character image into an image encoder of an image encryption model to obtain the characteristics of the keyboard character image, and inputting the wrong keyboard character into a character encoder of the image encryption model to obtain the characteristics of the wrong keyboard character;
and fusing the characteristics of the keyboard character image and the characteristics of the wrong keyboard character and inputting the fused keyboard character image and the fused keyboard character image into an image decoder of the image encryption model to obtain the encrypted keyboard character image.
3. The method of claim 2, wherein the training process of the image encryption model comprises:
training the image encoder and the image decoder by taking the same visual recognition results of the original character image and the encrypted character image as a model training target;
training the character encoder by taking the neural network recognition result of the encrypted character image as an error character as a model training target; the error character is different from the character corresponding to the original character image.
4. A computer device, comprising:
the encryption unit is used for extracting the characteristics of the keyboard character image, determining fusion characteristics according to the characteristics of the keyboard character image and the characteristics of wrong keyboard characters, and determining the encrypted keyboard character image according to the fusion characteristics; the wrong keyboard character is different from the character corresponding to the keyboard character image;
the construction unit is used for constructing an encrypted keyboard based on the encrypted keyboard character image; and the neural network recognition result of the encrypted keyboard character image is the wrong keyboard character, and the visual recognition results of the encrypted keyboard character image and the keyboard character image are the same.
5. The computer device of claim 4,
the encryption unit is specifically configured to input the keyboard character image into an image encoder of an image encryption model to obtain the feature of the keyboard character image, and input the incorrect keyboard character into a character encoder of the image encryption model to obtain the feature of the incorrect keyboard character; and fusing the characteristics of the keyboard character image and the characteristics of the wrong keyboard character and inputting the fused keyboard character image and the fused keyboard character image into an image decoder of the image encryption model to obtain the encrypted keyboard character image.
6. The computer device of claim 5, wherein the training process of the image encryption model comprises:
training the image encoder and the image decoder by taking the same visual recognition results of the original character image and the encrypted character image as a model training target;
training the character encoder by taking the neural network recognition result of the encrypted character image as an error character as a model training target; the error character is different from the character corresponding to the original character image.
7. A server comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 3 are implemented by the processor when executing the computer program.
8. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 3.
9. A computer program product comprising a computer program, characterized in that the computer program realizes the steps of the method of any one of claims 1-3 when executed by a processor.
10. A method for displaying an encrypted keyboard character image generated by the method of any one of claims 1-3.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110069624.9A CN114817937A (en) | 2021-01-19 | 2021-01-19 | Keyboard encryption method, device, storage medium and computer program product |
| PCT/CN2022/072040 WO2022156609A1 (en) | 2021-01-19 | 2022-01-14 | Keyboard encryption method, and device, storage medium and computer program product |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110069624.9A CN114817937A (en) | 2021-01-19 | 2021-01-19 | Keyboard encryption method, device, storage medium and computer program product |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114817937A true CN114817937A (en) | 2022-07-29 |
Family
ID=82525010
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110069624.9A Pending CN114817937A (en) | 2021-01-19 | 2021-01-19 | Keyboard encryption method, device, storage medium and computer program product |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN114817937A (en) |
| WO (1) | WO2022156609A1 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114817937A (en) * | 2021-01-19 | 2022-07-29 | 北京嘀嘀无限科技发展有限公司 | Keyboard encryption method, device, storage medium and computer program product |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103996011A (en) * | 2014-06-05 | 2014-08-20 | 福建天晴数码有限公司 | Method and device for protecting codes to be input safely |
| CN107360137A (en) * | 2017-06-15 | 2017-11-17 | 深圳市牛鼎丰科技有限公司 | Construction method and device for the neural network model of identifying code identification |
| CN108446700A (en) * | 2018-03-07 | 2018-08-24 | 浙江工业大学 | A kind of car plate attack generation method based on to attack resistance |
| US20190114310A1 (en) * | 2015-05-04 | 2019-04-18 | King Fahd University Of Petroleum And Minerals | Method for analyzing and reproducing text data |
| CN109993162A (en) * | 2019-03-01 | 2019-07-09 | 昆明理工大学 | Laotian block letter text optical character recognition methods based on convolutional neural networks |
| CN110866238A (en) * | 2019-11-13 | 2020-03-06 | 北京工业大学 | Generation method of verification code image based on countermeasure sample |
| CN111325319A (en) * | 2020-02-02 | 2020-06-23 | 腾讯云计算（北京）有限责任公司 | Method, device, equipment and storage medium for detecting neural network model |
| CN111460837A (en) * | 2020-03-31 | 2020-07-28 | 广州大学 | Character-level confrontation sample generation method and device for neural machine translation |
| CN111475797A (en) * | 2020-03-26 | 2020-07-31 | 深圳先进技术研究院 | Method, device and equipment for generating confrontation image and readable storage medium |
| CN111507093A (en) * | 2020-04-03 | 2020-08-07 | 广州大学 | Text attack method and device based on similar dictionary and storage medium |
| WO2022156609A1 (en) * | 2021-01-19 | 2022-07-28 | 北京嘀嘀无限科技发展有限公司 | Keyboard encryption method, and device, storage medium and computer program product |
-
2021
- 2021-01-19 CN CN202110069624.9A patent/CN114817937A/en active Pending
-
2022
- 2022-01-14 WO PCT/CN2022/072040 patent/WO2022156609A1/en active Application Filing
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103996011A (en) * | 2014-06-05 | 2014-08-20 | 福建天晴数码有限公司 | Method and device for protecting codes to be input safely |
| US20190114310A1 (en) * | 2015-05-04 | 2019-04-18 | King Fahd University Of Petroleum And Minerals | Method for analyzing and reproducing text data |
| CN107360137A (en) * | 2017-06-15 | 2017-11-17 | 深圳市牛鼎丰科技有限公司 | Construction method and device for the neural network model of identifying code identification |
| CN108446700A (en) * | 2018-03-07 | 2018-08-24 | 浙江工业大学 | A kind of car plate attack generation method based on to attack resistance |
| CN109993162A (en) * | 2019-03-01 | 2019-07-09 | 昆明理工大学 | Laotian block letter text optical character recognition methods based on convolutional neural networks |
| CN110866238A (en) * | 2019-11-13 | 2020-03-06 | 北京工业大学 | Generation method of verification code image based on countermeasure sample |
| CN111325319A (en) * | 2020-02-02 | 2020-06-23 | 腾讯云计算（北京）有限责任公司 | Method, device, equipment and storage medium for detecting neural network model |
| CN111475797A (en) * | 2020-03-26 | 2020-07-31 | 深圳先进技术研究院 | Method, device and equipment for generating confrontation image and readable storage medium |
| CN111460837A (en) * | 2020-03-31 | 2020-07-28 | 广州大学 | Character-level confrontation sample generation method and device for neural machine translation |
| CN111507093A (en) * | 2020-04-03 | 2020-08-07 | 广州大学 | Text attack method and device based on similar dictionary and storage medium |
| WO2022156609A1 (en) * | 2021-01-19 | 2022-07-28 | 北京嘀嘀无限科技发展有限公司 | Keyboard encryption method, and device, storage medium and computer program product |
Non-Patent Citations (2)
| Title |
|---|
| SOURABH VASANT GOTHE等: "Language Detection Engine for Multilingual Texting on Mobile Devices", 《2020 IEEE 14TH INTERNATIONAL CONFERENCE ON SEMANTIC COMPUTING (ICSC)》, 12 March 2020 (2020-03-12), pages 279 - 286 * |
| 杨莉等: "基于数据融合的车内空气质量控制系统的研究", 《机电一体化》, vol. 15, no. 05, 20 May 2009 (2009-05-20), pages 46 - 49 * |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2022156609A1 (en) | 2022-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11122089B2 (en) | Authorization policy optimization method and apparatus, and storage medium | |
| US11995155B2 (en) | Adversarial image generation method, computer device, and computer-readable storage medium | |
| CN108875404B (en) | Data desensitization method and device based on machine learning and storage medium | |
| CN109523611B (en) | Verification code picture generation method and device | |
| CN110414567A (en) | Data processing method, device and electronic equipment | |
| US10778716B2 (en) | Detecting webpages that share malicious content | |
| US11847210B2 (en) | Detecting device and detecting method | |
| EP3830723B1 (en) | Increasing security of a password-protected resource based on publicly available data | |
| CN113762328B (en) | Model training method, device, equipment and storage medium based on federal learning | |
| CN113240430B (en) | Mobile payment verification method and device | |
| CN112397057A (en) | Voice processing method, device, equipment and medium based on generation countermeasure network | |
| CN112398838A (en) | Authentication method, device and equipment based on privacy protection | |
| CN114266336B (en) | Method for processing artificial intelligent model, and method and device for processing data | |
| CN111382403A (en) | Training method, device, equipment and storage medium of user behavior recognition model | |
| CN114817937A (en) | Keyboard encryption method, device, storage medium and computer program product | |
| CN110519056A (en) | A kind of login method, method for generating cipher code and its relevant apparatus | |
| EP3734484A1 (en) | System for providing query processing service on basis of personal information protection | |
| CN107623664A (en) | A kind of cipher-code input method and device | |
| CN113329004A (en) | Authentication method, system and device | |
| CN114817893A (en) | Authentication code image encryption method, device, storage medium and computer program product | |
| CN114387480A (en) | Portrait scrambling method and device | |
| CN110348185B (en) | Method, device and equipment for realizing verification code and computer storage medium | |
| CN113961962A (en) | Model training method and system based on privacy protection and computer equipment | |
| CN114004206A (en) | Form generation method and device, computer equipment and storage medium | |
| CN113674083A (en) | Internet financial platform credit risk monitoring method, device and computer system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |