CN111666968A - Man-machine recognition method and device, electronic equipment and computer readable storage medium - Google Patents

Man-machine recognition method and device, electronic equipment and computer readable storage medium Download PDF

Info

Publication number
CN111666968A
CN111666968A CN202010319326.6A CN202010319326A CN111666968A CN 111666968 A CN111666968 A CN 111666968A CN 202010319326 A CN202010319326 A CN 202010319326A CN 111666968 A CN111666968 A CN 111666968A
Authority
CN
China
Prior art keywords
preset
data
target
human
operation track
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
Application number
CN202010319326.6A
Other languages
Chinese (zh)
Inventor
周开波
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing QIYI Century Science and Technology Co Ltd
Original Assignee
Beijing QIYI Century Science and Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing QIYI Century Science and Technology Co Ltd filed Critical Beijing QIYI Century Science and Technology Co Ltd
Priority to CN202010319326.6A priority Critical patent/CN111666968A/en
Publication of CN111666968A publication Critical patent/CN111666968A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
    • 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
    • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/31—User authentication
    • 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
    • G06N3/045—Combinations of networks
    • 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
    • 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
    • G06F2221/2133—Verifying human interaction, e.g., Captcha

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Security & Cryptography (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computer Hardware Design (AREA)
  • Image Analysis (AREA)

Abstract

The embodiment of the invention provides a man-machine identification method and device, electronic equipment and a computer readable storage medium, and relates to the technical field of information security. The man-machine identification method comprises the following steps: preprocessing target operation track data to generate input data of a preset human-computer recognition model; generating a target characterization vector corresponding to the input data through the preset human-computer recognition model; and determining the type of the target operation track data according to the position relation of the target characterization vector and a plurality of preset characterization vectors in a preset space. The technical scheme provided by the embodiment of the invention solves the problems of poor generalization performance and certain hysteresis in the man-machine identification mode aiming at the attack characteristics in the prior art to a certain extent.

Description

Man-machine recognition method and device, electronic equipment and computer readable storage medium
Technical Field
The present invention relates to the field of information security technologies, and in particular, to a human-machine recognition method and apparatus, an electronic device, and a computer-readable storage medium.
Background
Human-machine identification is a public turing test used to identify whether an operator is a real person or a machine. In the prior art, in application scenarios such as registration, login, voting, content upload, etc., various verification methods (such as sliding verification or word selection verification, etc.) can be used for man-machine identification, so as to avoid an attacker from performing verification code input by simulating a real person through a machine as much as possible, attack website services, and reduce the occurrence of malicious behaviors such as swiping a ticket, weeding wool, mass upload of low-quality content, etc.
The existing man-machine identification technology mainly aims at defending attack characteristics, namely, mainly identifies machine fake tracks. However, the attacker can change the characteristics of the fake track by adding smoothing processing to the fake track data and adding random signals to the fake track data, adding redundant tracks, and the like, to continuously circumvent the defense. After the characteristics of the forged track change, the original man-machine recognition model for man-machine recognition is difficult to accurately recognize the new forged track, and needs to learn again aiming at the new forged track characteristics, so that the defense mode aiming at the attack characteristics has poor generalization performance and certain hysteresis.
Disclosure of Invention
The invention provides a man-machine identification method and device, electronic equipment and a computer readable storage medium, which are used for solving the problems of poor generalization performance and certain hysteresis in a man-machine identification mode aiming at attack characteristics in the prior art to a certain extent.
In a first aspect of the present invention, there is provided a human-machine identification method applied to a server, the human-machine identification method including:
preprocessing target operation track data to generate input data of a preset human-computer recognition model;
generating a target characterization vector corresponding to the input data through the preset human-computer recognition model;
determining the type of the target operation track data according to the position relation of the target characterization vector and a plurality of preset characterization vectors in a preset space;
wherein the preset characterization vector is: generating a characterization vector by the preset human-computer recognition model according to the real human operation track data; the types of operation trajectory data include: real person operation track data and machine operation track data.
In a second aspect of the present invention, there is provided a human-machine recognition apparatus applied to a server, the human-machine recognition apparatus including:
the first generation module is used for preprocessing the target operation track data and generating input data of a preset human-computer recognition model;
the second generation module is used for generating a target characterization vector corresponding to the input data through the preset human-computer recognition model;
the type determining module is used for determining the type of the target operation track data according to the position relation of the target characterization vector and a plurality of preset characterization vectors in a preset space;
wherein the preset characterization vector is: generating a characterization vector by the preset human-computer recognition model according to the real human operation track data; the types of operation trajectory data include: real person operation track data and machine operation track data.
In a third aspect of the present invention, there is also provided an electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete communication with each other through the communication bus;
a memory for storing a computer program;
the processor is used for realizing the steps of the man-machine identification method when the program stored in the memory is executed.
In a fourth aspect implemented by the present invention, there is also provided a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the human recognition method as described above.
In a fifth aspect of the embodiments of the present invention, there is also provided a computer program product containing instructions which, when run on a computer, cause the computer to perform the human recognition method as described above.
Aiming at the prior art, the invention has the following advantages:
in the embodiment of the invention, when the man-machine recognition is carried out, the type of the operation track to be verified (namely the target operation track data) is recognized according to the real person operation track (namely the preset characterization vector corresponding to the real person operation track data), so that no matter how an attacker changes the fake operation track, as long as the operation track does not accord with the characteristics of the real person operation track, the operation track can be regarded as the machine operation track. The man-machine identification method is not influenced by the change of the fake operation track, overcomes the hysteresis of a defense mode depending on the characteristics of the fake operation track, has better generalization performance and is beneficial to enhancing the man-machine identification capability.
The foregoing description is only an overview of the technical solutions of the present invention, and the embodiments of the present invention are described below in order to make the technical means of the present invention more clearly understood and to make the above and other objects, features, and advantages of the present invention more clearly understandable.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments will be briefly described below.
Fig. 1 is a schematic flow chart of a human-machine recognition method according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart illustrating another human-machine recognition method according to an embodiment of the present invention;
FIG. 3 is a diagram of a DNN self-encoder model according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating an RNN self-encoder model according to an embodiment of the present invention;
FIG. 5 is a flowchart illustrating another human-machine recognition method according to an embodiment of the present invention;
FIG. 6 is a schematic flow chart of model training according to an embodiment of the present invention;
FIG. 7 is a diagram illustrating a relationship between model training and human-machine recognition according to an embodiment of the present invention;
fig. 8 is a block diagram of a human-machine recognition device according to an embodiment of the present invention;
FIG. 9 is a block diagram of another human recognition device provided in an embodiment of the present invention;
FIG. 10 is a block diagram of another human recognition device provided in an embodiment of the present invention;
fig. 11 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
Fig. 1 is a schematic flow chart of a human-machine recognition method according to an embodiment of the present invention. The man-machine recognition method is applied to the server.
As shown in fig. 1, the human-machine recognition method may include:
step 101: and preprocessing the target operation track data to generate input data of a preset human-computer recognition model.
The target operation trajectory data is data generated in a man-machine verification process (namely, after verification is triggered and before verification is finished) and acquired by terminal equipment, and specifically may be sliding operation trajectory data and the like. The terminal equipment can send the target operation track data to the server after acquiring the target operation track data, so that the server obtains the target operation track data and carries out man-machine identification according to the target operation track data.
Generally, the operation trajectory data is a coordinate sequence. For example, when a user performs a mobile phone verification on a terminal device, the terminal device may capture a trajectory of a mouse or a finger sliding across a screen. Each time of man-machine verification, the sliding track can be sampled at n moments to obtain a coordinate sequence comprising n coordinates, such as: (x)1,y1,t1)、(x2,y2,t2)、…、(xi,yi,ti)、…、(xn,yn,tn). Wherein x isiX-seat of screen pointed by mouse or finger at ith momentTarget pixel value, yiY-coordinate pixel value, t, representing the screen pointed by the mouse or finger at time iiIndicating the timestamp corresponding to the ith time.
In the embodiment of the invention, the server adopts a pre-trained human-computer recognition model (namely a preset human-computer recognition model) to perform human-computer recognition according to the target operation track data. Before the target operation trajectory data is identified, corresponding preprocessing is required to be performed on the target operation trajectory data to obtain input data capable of being input into the preset human-machine identification model. The preprocessing mode may be different for different types of human-computer recognition models, and the specific situation may be selected according to actual requirements, which is not limited in the embodiments of the present invention.
Step 102: and generating a target characterization vector corresponding to the input data by presetting a human-computer recognition model.
And after preprocessing the target operation track data by the server to obtain input data, inputting the input data into a preset human-computer recognition model. The preset human-machine recognition model can generate a characterization vector (i.e. a target characterization vector) corresponding to the input data according to the input data.
In general, the dimension of the characterization vector generated by the preset human-machine recognition model is lower than that of the input data. The token vector is generated from the input data, which may be understood as a refinement of the input data to dimension-reduce the input data into a token vector. The characterization vector contains the most essential information of the operation trajectory, i.e. the operation trajectory can be restored according to the characterization vector, and therefore, the characterization vector can be used to represent the operation trajectory.
Step 103: and determining the type of the target operation track data according to the position relation of the target characterization vector and a plurality of preset characterization vectors in a preset space.
The dimensions of the preset space described herein are the same as the dimensions of the target characterization vector. The parameters represented by each dimension of the preset space are parameters corresponding to each element in the target characterization vector.
The preset characterization vector is as follows: the method comprises the following steps that a representation vector generated by a preset human-computer recognition model according to real human operation track data specifically comprises the following steps: and generating a characterization vector according to input data obtained by preprocessing the real person operation track data. The preset characterization vectors are multiple in number, and the specific number is targeted to include the characterization vectors corresponding to various human operation tracks as much as possible.
The types of operation trajectory data described herein include: real person operation track data and machine operation track data. The operator corresponding to the real-person operation track data is a real person, and the operator corresponding to the machine operation track data is a machine.
For the characterization vectors generated according to the real-person operation trajectory data, the corresponding spatial points in the preset space are often distributed in a certain area in a centralized manner, and therefore, the distribution of the spatial points corresponding to the preset characterization vectors in the preset space represents the possible distribution of the real-person trajectory in the preset space.
In the preset space, the target representation vector is closer to the preset representation vector, and the possibility that the target operation track data belongs to the real person operation track data is higher; conversely, the farther the target characterization vector is away from the preset characterization vector, the greater the possibility that the target operation trajectory data belongs to the machine operation trajectory data is.
The method for judging the position relationship between the target characterization vector and the plurality of preset characterization vectors in the preset space can be set according to actual requirements, and the method is not limited in the embodiment of the invention.
In the embodiment of the invention, when the man-machine recognition is carried out, the type of the operation track to be verified (namely the target operation track data) is recognized according to the real person operation track (namely the preset characterization vector corresponding to the real person operation track data), so that no matter how an attacker changes the fake operation track, as long as the operation track does not accord with the characteristics of the real person operation track, the operation track can be regarded as the machine operation track. The man-machine identification method is not influenced by the change of the fake operation track, overcomes the hysteresis of a defense mode depending on the characteristics of the fake operation track, has better generalization performance and is beneficial to enhancing the man-machine identification capability.
Fig. 2 is a flowchart illustrating another man-machine recognition method according to an embodiment of the present invention. The man-machine recognition method is applied to the server.
As shown in fig. 2, the human-machine recognition method may include:
step 201: and preprocessing the target operation track data to generate input data of a preset human-computer recognition model.
For the explanation of this step, reference may be made to the foregoing detailed description of step 101, and thus will not be described herein again.
Step 202: and generating a target characterization vector corresponding to the input data by presetting a human-computer recognition model.
For the explanation of this step, reference may be made to the foregoing detailed description of step 102, and further description is omitted here.
Step 203: and determining whether the target space point in the preset space has the preset space point within a preset range.
The dimensions of the preset space described herein are the same as the dimensions of the target characterization vector. The parameters represented by each dimension of the preset space are parameters corresponding to each element in the target characterization vector.
The target space point is a corresponding space point of the target characterization vector in a preset space. The preset spatial point is a corresponding spatial point of the preset characterization vector in the preset space. For example, an m-dimensional token vector may be represented as (f)1,f2,…,fm) And may also be represented as a point in m-dimensional space.
The preset characterization vector is as follows: the method comprises the following steps that a representation vector generated by a preset human-computer recognition model according to real human operation track data specifically comprises the following steps: and generating a characterization vector according to input data obtained by preprocessing the real person operation track data. The preset characterization vectors are multiple in number, and the specific number is targeted to include the characterization vectors corresponding to various human operation tracks as much as possible.
For the characterization vectors generated according to the real-person operation trajectory data, the corresponding spatial points in the preset space are often distributed in a certain area in a concentrated manner, and therefore, the distribution of the spatial points corresponding to the preset characterization vectors in the preset space represents the possible distribution of the real-person trajectory in the preset space, and therefore, in the embodiment of the invention, whether the target characterization vectors are in the distribution area of the real-person trajectory can be determined by judging whether the spatial points corresponding to the preset characterization vectors exist within a certain euclidean distance (i.e., within a preset range) of the corresponding spatial points in the preset space, and further, whether the operation trajectory data (i.e., the target operation trajectory data) corresponding to the target characterization vectors is the real-person operation trajectory data or the machine operation trajectory data is determined. Wherein, the smaller the preset range is set, the more accurate the obtained judgment result is. Of course, the more the number of the preset characterization vectors is, the more accurate the obtained judgment result is.
Alternatively, it may be determined whether there is a preset spatial point within a preset range of the target spatial point in the preset space by a spatial proximity algorithm (e.g., K-Nearest Neighbor (KNN) algorithm).
Step 204: and under the condition that the target space point has the preset space point in the preset range, determining that the target operation track data is the real person operation track data.
When it is determined that the spatial point corresponding to the target characterization vector is within a certain euclidean distance and the spatial point corresponding to the preset characterization vector exists, it is determined that the target characterization vector is in a distribution area of the real person trajectory, and the operation trajectory data corresponding to the target characterization vector is considered to be the real person operation trajectory data, that is, the operator corresponding to the target operation trajectory data is the real person.
Step 205: and under the condition that the preset space point does not exist in the preset range of the target space point, determining the target operation track data as machine operation track data.
Under the condition that it is determined that the spatial point corresponding to the target characterization vector does not have the spatial point corresponding to the preset characterization vector within a certain Euclidean distance, if the target characterization vector has a larger distribution space which is probably not in the trajectory of a real person in the preset space, the operation trajectory data corresponding to the target characterization vector can be regarded as machine operation trajectory data, that is, the operator corresponding to the target operation trajectory data is a machine.
In the embodiment of the invention, when the man-machine recognition is carried out, the type of the operation track to be verified (namely the target operation track data) is recognized according to the real person operation track (namely the preset characterization vector corresponding to the real person operation track data), so that no matter how an attacker changes the fake operation track, as long as the operation track does not accord with the characteristics of the real person operation track, the operation track can be regarded as the machine operation track. The man-machine identification method is not influenced by the change of the fake operation track, overcomes the hysteresis of a defense mode depending on the characteristics of the fake operation track, has better generalization performance and is beneficial to enhancing the man-machine identification capability.
Optionally, the preset human-machine recognition model in the embodiment of the present invention may be a self-encoder model. The self-encoder is an artificial neural network model used in semi-supervised learning and unsupervised learning, and can be used for performing characterization learning on input information by taking the input information as a learning target.
In general, the self-coder model includes: an encoder (i.e., encoder) and a decoder (i.e., decoder).
The encoder is responsible for mapping the original input features into a token vector, i.e. the output of the encoder is a token vector. When the preset human-computer recognition model is used for human-computer recognition, the characterization vector output by the encoder is a required result. In the embodiment of the invention, the input data corresponding to the target operation track data is input information of an encoder, and the target characterization vector is output information of the encoder.
The decoder is mainly used in the model training process, and is responsible for decoding the characterization vector output by the encoder and trying to restore the characterization vector to the original input features, and the training optimization goal of the self-encoder model is as follows: the reconstruction error of the original input features is minimized, i.e. the error between the output data of the decoder and the input data of the encoder is reduced.
Optionally, the artificial neural network may further include: deep Neural Networks (DNNs) and Recurrent Neural Networks (RNNs). The self-encoder in the embodiment of the present invention may be constructed by using the two artificial neural networks, respectively, so that the self-encoder model may further include: a deep neural network self-encoder (hereinafter referred to as a DNN self-encoder) and a recurrent neural network self-encoder (hereinafter referred to as an RNN self-encoder).
(1) The encoder part and the decoder part of the DNN self-encoder are both composed of a plurality of layers of fully-connected neural networks, and the input layer of the DNN self-encoder is composed of neurons with the same number as that of input characteristic parameters. Assuming that the number of input feature parameters is 3, the number of input layer neurons of the DNN self-encoder is also 3, and correspondingly, the number of output layer neurons of the decoder is also 3. For the hidden layers of the DNN self-encoder, the number of neurons can be set according to actual requirements, and the number of the hidden layer neurons is generally smaller than the number of input neurons and output neurons, as shown in fig. 3, a circle with a diagonal filling pattern in the figure represents a hidden layer neuron, and the number of neurons in each hidden layer is 2.
For the deep neural network, the operation trajectory data cannot be directly input into the model composed of the operation trajectory data, and the feature data extraction needs to be performed according to the preset feature parameters, so that, step 201: preprocessing the target operation trajectory data to generate input data of a preset human-computer recognition model, which may include:
the method comprises the following steps: and acquiring a data sequence corresponding to each preset parameter according to the target operation track data.
The preset parameters described herein include: at least one of speed, acceleration, and a time interval between two adjacent sampling times when the operation trajectory data is collected.
The target operation track data is a coordinate sequence, and each coordinate in the coordinate sequence is data acquired by the operation track at each sampling moment. According to the coordinate sequence, the speed and the acceleration of each sampling moment and the time interval between two adjacent sampling moments can be calculated. For the parameters speed, acceleration and time interval, a data sequence can be obtained, for example, for the speed parameter, a speed sequence can be obtained: v. of1、v2、…、vn。
Step two: and acquiring target statistics corresponding to each data sequence.
Target statistics as described herein include: at least one of a maximum value in the data sequence, a minimum value in the data sequence, an average value of the data sequence, and an information entropy of the data sequence.
After the data sequence of each preset parameter is obtained, the maximum value and the minimum value in each data sequence, and the statistics of the average value, the variance, the information entropy, and the like of the data sequence can be calculated. For example, statistics such as a maximum speed value, a minimum speed value, a mean speed value, and a variance speed can be obtained from a speed sequence corresponding to a speed parameter.
Step three: and determining the obtained target statistic as input data of a preset man-machine recognition model.
In the embodiment of the invention, the acquired target statistic can be used as the input data of the preset human-machine recognition model.
(2) The encoder part and the decoder part of the RNN self-encoder are both composed of a recurrent neural network. As illustrated in fig. 4, the input data of the RNN self-encoder may be a sequence, namely: upsilon is1、υ2、υ3(ii) a The output data of the RNN self-encoder from the decoder is also a sequence, namely: upsilon is1'、υ2'、υ3'. Wherein w in the figure1For encoder parameters, w in the figure2Are decoder parameters.
For the cyclic neural network, after normalization processing is carried out on target operation track data, the target operation track data can be used as input data of a model formed by the target operation track data.
The normalization methods used in the normalization process may include, but are not limited to: maximum-minimum normalization (min-maxscaling) and 0-mean normalization (z-score normalization).
Assuming that the coordinate sequence corresponding to the target operation track data is as follows: (x)1,y1,t1)、(x2,y2,t2)、…、(xi,yi,ti)、…、(xn,yn,tn) The two normalization methods are explained below by taking the x parameter as an example.
Maximum and minimum normalization
The formula: x is the number ofnorm＝(xi-xmin)/(xmax-xmin)。
Wherein x isnormDenotes the normalized x value, xiRepresenting the original x-value, x, in a coordinate sequenceminRepresenting the minimum value of the x parameter, x, in a coordinate sequencemaxRepresenting the maximum value of the x parameter in the coordinate sequence.
When the maximum and minimum normalization processing is performed on the original x values in the coordinate sequence, the normalization processing may be performed on each original x value in sequence by using the above formula.
0 mean value normalization
The formula: z ═ xi-μ)/σ。
Wherein z represents the normalized x value, xiRepresenting the original x value in the coordinate sequence, mu represents the mean of the x parameter in the coordinate sequence, xmaxRepresenting the variance of the x parameter in the coordinate sequence.
When the 0-mean normalization is performed on the original x values in the coordinate sequence, the above formula may be used to sequentially perform normalization on each original x value.
In summary, in the embodiment of the present invention, when performing human-machine recognition, the type of the operation track to be verified (i.e., the target operation track data) is recognized according to the real-person operation track (i.e., the preset characterization vector corresponding to the real-person operation track data). Specifically, whether the target characterization vector is in a distribution area of a real person track is determined by judging whether the target characterization vector exists within a certain Euclidean distance (namely, within a preset range) of a corresponding space point in a preset space or not, and further, whether operation track data (namely, target operation track data) corresponding to the target characterization vector is real person operation track data or machine operation track data is determined. Therefore, no matter how the attacker changes the fake operation track, as long as the characteristic of the real person operation track is not met, the operation track to be detected can be regarded as the machine operation track without being influenced by the change of the fake operation track, the hysteresis of a defense mode depending on the characteristic of the fake operation track is overcome, the better generalization performance is realized, and the capability of human-computer identification is favorably enhanced.
Fig. 5 is a flowchart illustrating another man-machine recognition method according to an embodiment of the present invention.
As shown in fig. 5, the human-machine recognition method may include:
step 501: and preprocessing the target operation track data to generate input data of a preset human-computer recognition model.
For the explanation of this step, reference may be made to the foregoing detailed description of step 101, and thus will not be described herein again.
Step 502: and generating a target characterization vector corresponding to the input data by presetting a human-computer recognition model.
For the explanation of this step, reference may be made to the foregoing detailed description of step 102, and further description is omitted here.
Step 503: and determining whether the target space point in the preset space is in a preset area in the preset space.
The dimensions of the preset space described herein are the same as the dimensions of the target characterization vector. The parameters represented by each dimension of the preset space are parameters corresponding to each element in the target characterization vector.
The target space point is a corresponding space point of the target characterization vector in a preset space.
The preset areas described here are: and the spatial range is determined in advance according to the plurality of preset characterization vectors. The preset characterization vector is as follows: the method comprises the following steps that a representation vector generated by a preset human-computer recognition model according to real human operation track data specifically comprises the following steps: and generating a characterization vector according to input data obtained by preprocessing the real person operation track data. The preset characterization vectors are multiple in number, and the specific number is targeted to include the characterization vectors corresponding to various human operation tracks as much as possible.
For the characterization vectors generated according to the real-person operation trajectory data, the corresponding spatial points in the preset space are often distributed in a certain area in a centralized manner, and therefore, the distribution of the spatial points corresponding to the preset characterization vectors in the preset space represents the possible distribution of the real-person trajectory in the preset space, and therefore, in the embodiment of the invention, whether the target characterization vectors are in the distribution area of the real-person trajectory can be determined by judging whether the spatial points corresponding to the target characterization vectors in the preset space are in the spatial range determined according to the spatial points corresponding to the preset characterization vectors, and further, whether the operation trajectory data (i.e., the target operation trajectory data) corresponding to the target characterization vectors is the real-person operation trajectory data or the machine operation trajectory data is determined. The more the number of the preset characterization vectors is, the more accurate the obtained preset region is, and the more accurate the judgment result is.
Optionally, when determining whether the target spatial point is within a preset region in a preset space, a corresponding spatial coordinate system may be established in the preset space, the target spatial point and the preset region are mapped in the spatial coordinate system, and then it is determined whether a coordinate point corresponding to the target spatial point is within a coordinate range corresponding to the preset region.
Step 504: and under the condition that the target space point is in the preset area, determining the operation track data as real person operation track data.
Under the condition that the spatial point corresponding to the target representation vector is determined to be in the preset area, the target representation vector is determined to be in the distribution area of the real person track, and then the operation track data corresponding to the target representation vector can be regarded as real person operation track data, namely, the operator corresponding to the target operation track data is a real person.
Step 505: and under the condition that the target space point is not in the preset area, determining the operation track data as machine operation track data.
Under the condition that the spatial point corresponding to the target representation vector is determined not to be in the preset area, the target representation vector is determined not to be in the distribution area of the real human trajectory, and then the operation trajectory data corresponding to the target representation vector can be regarded as machine operation trajectory data, that is, the operator corresponding to the target operation trajectory data is a machine.
In the embodiment of the invention, when man-machine recognition is carried out, the type of the operation track to be verified (namely target operation track data) is recognized according to the real person operation track (namely the preset characterization vector corresponding to the real person operation track data), so that no matter how an attacker changes the fake operation track, as long as the characteristic of the real person operation track is not met, the operation track to be detected can be regarded as the machine operation track without being influenced by the change of the fake operation track, the hysteresis of a mode of defending by depending on the characteristic of the fake operation track is overcome, better generalization performance is realized, and the man-machine recognition capability is favorably enhanced.
Optionally, the preset human-machine recognition model in the embodiment of the present invention may be a self-encoder model, and for the explanation of the self-encoder, reference may be made to the description of the self-encoder, and in order to avoid repetition, details are not repeated here.
Optionally, the artificial neural network may further include: deep neural networks and recurrent neural networks. The self-encoder in the embodiment of the present invention may be constructed by using the two artificial neural networks, respectively, so that the self-encoder model may further include: a deep neural network self-encoder (hereinafter referred to as a DNN self-encoder) and a recurrent neural network self-encoder (hereinafter referred to as an RNN self-encoder). For a detailed explanation of the DNN self-encoder and the RNN self-encoder, reference may be made to the foregoing description for the two, and details are not repeated here to avoid repetition.
In summary, in the embodiment of the present invention, when performing human-machine recognition, the type of the operation track to be verified (i.e., the target operation track data) is recognized according to the real-person operation track (i.e., the preset characterization vector corresponding to the real-person operation track data). Specifically, whether the target representation vector is in the distribution area of the real person track is determined by judging whether a space point corresponding to the target representation vector in a preset space is in a space range determined according to the space point corresponding to the preset representation vector, and then whether operation track data (namely target operation track data) corresponding to the target representation vector is real person operation track data or machine operation track data is determined. Therefore, no matter how the attacker changes the fake operation track, as long as the characteristic of the real person operation track is not met, the operation track to be detected can be regarded as the machine operation track without being influenced by the change of the fake operation track, the hysteresis of a defense mode depending on the characteristic of the fake operation track is overcome, the better generalization performance is realized, and the capability of human-computer identification is favorably enhanced.
Fig. 6 is a flowchart illustrating a training method of a human-machine recognition model according to an embodiment of the present invention. The training method is suitable for the man-machine recognition model with the model type of the self-encoder.
As shown in fig. 6, the training method of the human-machine recognition model may include:
step 601: and preprocessing a first group of sample data in the multiple groups of sample data to generate first input sample data.
As shown in fig. 7, before training a preset human-computer recognition model, a certain amount of real-person operation trajectory data is collected, and then the real-person operation trajectory data is preprocessed for training the human-computer recognition model. The number of the real person operation trajectory data can be specifically set according to actual requirements, and the embodiment of the invention does not limit the number.
Wherein, each group of operation track data is a coordinate sequence, namely: (x)1,y1,t1)、(x2,y2,t2)、…、(xi,yi,ti)、…、(xn,yn,tn). Wherein x isiX-coordinate pixel value, y, representing the screen pointed by the mouse or finger at time iiY-coordinate pixel value, t, representing the screen pointed by the mouse or finger at time iiIndicating the timestamp corresponding to the ith time.
After the sample data collection is completed, each group of sample data is used to train the preset human-machine recognition model one by one, and the following description will take the first group of sample data as an example to explain the training process.
For example, a first set of sample data is preprocessed to become input data (i.e. first input sample data) capable of being input into a preset human-machine recognition model. The specific preprocessing mode can be determined according to the type of the artificial neural network forming the human-computer recognition model, and the preprocessing modes adopted by the deep neural network and the recurrent neural network are taken as examples and explained, and are not repeated here.
Step 602: and inputting the first input sample data into an encoder of a preset human-computer recognition model to obtain a first characterization vector corresponding to the first input sample data.
And after the obtained first input sample data is input into an encoder of a preset human-computer recognition model, and a first characterization vector is generated by the encoder according to the first input sample data.
Step 603: and inputting the first characterization vector into a decoder of a preset human-computer recognition model to obtain first output data corresponding to the first characterization vector.
And after the first characterization vector is obtained, inputting the first characterization vector into a decoder as input data of the decoder of the preset human-machine recognition model. First output data is generated by a decoder from the first token vector.
Step 604: an error value between the first output data and the first input sample data is determined.
After the first output data is obtained, the first output data is compared with first input sample data, and an error value of the first output data relative to the first input sample data is determined.
Alternatively, the Error value may specifically be a Mean Squared Error (MSE). The mean square error is the expected value of the square of the difference between the estimated value of the parameter and the true value of the parameter. The mean square error can evaluate the change degree of the data, and the smaller the mean square error value is, the better the accuracy of the prediction model describing the experimental data is.
Step 605: and adjusting the parameters of the preset human-computer recognition model under the condition that the error value is greater than the preset error value.
The smaller the error value is, the more similar the first output data and the first input sample data are, and the more accurate the output result of the preset man-machine recognition model is; on the contrary, the larger the error value is, the larger the difference between the first output data and the first input sample data is, the more the precision of the output result of the preset human-computer recognition model still needs to be improved, therefore, in the embodiment of the present invention, an error threshold (preset error value) may be set, so that the error value between the first output data and the first input sample data is compared with the preset error value, and under the condition that the error value is less than or equal to the preset error value, the prediction precision of the preset human-computer recognition model is considered to have reached the desired precision; when the error value is greater than the preset error value, the prediction accuracy of the preset human-computer recognition model is not yet achieved to the expected accuracy, parameters of the preset human-computer recognition model need to be adjusted, and the preset human-computer recognition model needs to be trained continuously to improve the model prediction accuracy.
Step 606: training the preset human-computer recognition model after the parameters are adjusted continuously by using the residual sample data until the training of the preset human-computer recognition model is determined to be finished under the condition that the error value between the output data of the decoder and the input sample data of the encoder is less than or equal to the preset error value for the first preset times; or determining to finish the training of the preset human-computer recognition model under the condition that the training times of the preset human-computer recognition model reach a second preset time.
After the parameters of the preset human-computer recognition model are adjusted once, the preset human-computer recognition model after the parameters are adjusted is continuously trained on the basis of the preset human-computer recognition model after the parameters are adjusted and the residual sample data.
When the error value between the output data of the decoder and the input sample data of the encoder is smaller than or equal to the preset error value, the prediction precision of the preset human-computer recognition model is stable and reaches the expected precision, and then the training of the preset human-computer recognition model is stopped; or when the training times of the preset human-computer recognition model reach a second preset time, which indicates that the training degree of the preset human-computer recognition model meets the requirement, stopping training the preset human-computer recognition model.
Alternatively, one or both of the above-described two training suspension conditions may be selected and used. Under the condition of all the conditions, which condition is met first, and the training of the preset human-computer recognition model is stopped based on which condition.
As shown in fig. 7, after the training of the preset human-machine recognition model is completed, a certain amount of real-person operation trajectory data may be generated by using the trained preset human-machine recognition model to generate real-person characterization vectors (corresponding to the preset characterization vectors described above), so as to form a real-person characterization vector space for subsequent human-machine recognition. It is understood that a preset number of human characterization vectors may also be randomly extracted from human characterization vectors generated from a certain number of human operation trajectory data for storage, and a human characterization vector space is formed by the randomly extracted human characterization vectors.
When the trained preset human-computer recognition model is used for human-computer recognition, the operation track data to be detected (corresponding to the target operation track data) needs to be preprocessed, wherein the preprocessing parameters are the same as the preprocessing parameters used when the preset human-computer recognition model is trained. Then, the preprocessed data is input into the preset human-machine recognition model for model prediction, and a characterization vector (corresponding to the target characterization vector) is generated. Then, judging whether the characterization vector belongs to a real person characterization vector space, and if so, determining that the operation track data to be detected is real person operation track data; and if not, determining that the operation track data to be detected is machine operation track data.
In the embodiment of the invention, the self-encoder model is used as the preset human-computer recognition model, and the operation track of the real person is learned in an unsupervised learning mode, so that the type of the operation track to be verified can be recognized according to the characteristics of the operation track of the real person during human-computer recognition. Since the man-machine recognition is performed by using the feature of the real-person operation track, no matter how the attacker changes the fake operation track, the operation track can be regarded as the machine operation track as long as the operation track does not conform to the feature of the real-person operation track. The man-machine identification method is not influenced by the change of the fake operation track, overcomes the hysteresis of a defense mode depending on the characteristics of the fake operation track, has better generalization performance and is beneficial to enhancing the man-machine identification capability.
Fig. 8 is a block diagram of a human-machine recognition device according to an embodiment of the present invention, which is applied to a server.
As shown in fig. 8, the human recognition device 800 may include:
the first generating module 801 is configured to preprocess the target operation trajectory data and generate input data of a preset human-machine recognition model.
A second generating module 802, configured to generate, through the preset human-machine recognition model, a target characterization vector corresponding to the input data.
A type determining module 803, configured to determine the type of the target operation trajectory data according to a position relationship between the target characterization vector and a plurality of preset characterization vectors in a preset space.
Wherein the preset characterization vector is: generating a characterization vector by the preset human-computer recognition model according to the real human operation track data; the types of operation trajectory data include: real person operation track data and machine operation track data.
In the embodiment of the invention, when man-machine recognition is carried out, the type of the operation track to be verified (namely target operation track data) is recognized according to the real person operation track (namely the preset characterization vector corresponding to the real person operation track data), so that no matter how an attacker changes the fake operation track, as long as the characteristic of the real person operation track is not met, the operation track to be detected can be regarded as the machine operation track without being influenced by the change of the fake operation track, the hysteresis of a mode of defending by depending on the characteristic of the fake operation track is overcome, better generalization performance is realized, and the man-machine recognition capability is favorably enhanced.
Optionally, the dimension of the preset space is the same as the dimension of the target characterization vector.
Optionally, as shown in fig. 9, the type determining module 803 includes:
the first determining unit 8031 is configured to determine whether there is a preset space point within a preset range of the target space point in the preset space.
The target space point is a corresponding space point of the target characterization vector in the preset space; the preset space point is a space point corresponding to the preset characterization vector in the preset space.
A first determining unit 8032, configured to determine that the target operation trajectory data is real operation trajectory data when the preset spatial point is included in the preset range of the target spatial point.
A second determining unit 8033, configured to determine that the target operation trajectory data is machine operation trajectory data when the preset spatial point is not included in the preset range of the target spatial point.
Optionally, as shown in fig. 10, the type determining module 803 includes:
a second determining unit 8034, configured to determine whether the target space point in the preset space is located in a preset area in the preset space.
The target space point is a corresponding space point of the target characterization vector in the preset space; the preset area is as follows: and determining a spatial range according to the plurality of preset characterization vectors.
A third determining unit 8035, configured to determine that the operation trajectory data is real-person operation trajectory data when the target space point is within the preset area.
A fourth determining unit 8036, configured to determine that the operation trajectory data is machine operation trajectory data when the target spatial point is not within the preset area.
Optionally, as shown in fig. 9 and 10, the human-machine recognition device 800 further includes:
and the model training module 804 is used for training the preset human-machine recognition model according to multiple groups of sample data.
And each group of the sample data is real person operation track data.
Optionally, the preset human-machine recognition model is a self-encoder model, and the self-encoder model includes: an encoder and a decoder.
As shown in fig. 9 and 10, in the case that the preset human-machine recognition model is an auto-encoder model, the model training module 804 includes:
a first generating unit 8041, configured to perform preprocessing on a first group of sample data in the multiple groups of sample data, and generate first input sample data.
A second generating unit 8042, configured to input the first input sample data into an encoder in the preset human-machine recognition model, and obtain a first characterization vector corresponding to the first input sample data.
A third generating unit 8043, configured to input the first token vector into a decoder in the preset hmi model, and obtain first output data corresponding to the first token vector.
A fifth determining unit 8044, configured to determine an error value between the first output data and the first input sample data.
A parameter adjusting unit 8045, configured to adjust a parameter of the preset human-machine recognition model when the error value is greater than a preset error value;
the processing unit 8046 is configured to continue training the preset human-machine recognition model after the parameters are adjusted by using the remaining sample data until an error value between output data of the decoder and input sample data of the encoder is smaller than or equal to a preset error value for a first preset number of consecutive times, and determine that training of the preset human-machine recognition model is completed; or determining to finish the training of the preset human-computer recognition model under the condition that the training times of the preset human-computer recognition model reach a second preset time.
Optionally, as shown in fig. 9 and 10, in a case that the neural network in the preset human-machine recognition model is a recurrent neural network, the first generating module 801 includes:
a fourth generating unit 8011, configured to perform normalization processing on the target operation trajectory data, and generate the input data.
Optionally, as shown in fig. 9 and 10, in the case that the neural network in the human-machine recognition model is a deep neural network model, the first generating module 801 includes:
the first obtaining unit 8012 is configured to obtain a data sequence corresponding to each preset parameter according to the target operation trajectory data.
Wherein the preset parameters include: at least one of speed, acceleration, and a time interval between two adjacent sampling times when the operation trajectory data is collected.
A second obtaining unit 8013 is configured to obtain a target statistic corresponding to each data sequence.
Wherein the target statistics include: at least one of a maximum value in the data sequence, a minimum value in the data sequence, an average value of the data sequence, and an information entropy of the data sequence.
A sixth determining unit 8014, configured to determine the acquired target statistic as the input data.
In the embodiment of the invention, when man-machine recognition is carried out, the type of the operation track to be verified (namely target operation track data) is recognized according to the real person operation track (namely the preset characterization vector corresponding to the real person operation track data), so that no matter how an attacker changes the fake operation track, as long as the characteristic of the real person operation track is not met, the operation track to be detected can be regarded as the machine operation track without being influenced by the change of the fake operation track, the hysteresis of a mode of defending by depending on the characteristic of the fake operation track is overcome, better generalization performance is realized, and the man-machine recognition capability is favorably enhanced.
For the above device embodiments, since they are basically similar to the method embodiments, reference may be made to the partial description of the method embodiments for relevant points.
The embodiment of the invention also provides the electronic equipment which can be a server. As shown in fig. 11, the system comprises a processor 1101, a communication interface 1102, a memory 1103 and a communication bus 1104, wherein the processor 1101, the communication interface 1102 and the memory 1103 are communicated with each other through the communication bus 1104.
A memory 1103 for storing a computer program.
When the electronic device is a terminal device, the processor 1101 is configured to execute the program stored in the memory 1103, and implement the following steps:
preprocessing target operation track data to generate input data of a preset human-computer recognition model;
generating a target characterization vector corresponding to the input data through the preset human-computer recognition model;
determining the type of the target operation track data according to the position relation of the target characterization vector and a plurality of preset characterization vectors in a preset space;
the dimension of the preset space is the same as that of the target characterization vector; the preset characterization vector is as follows: generating a characterization vector by the preset human-computer recognition model according to the real human operation track data; the types of operation trajectory data include: real person operation track data and machine operation track data.
Optionally, the determining the type of the target operation trajectory data according to the position relationship between the target characterization vector and a plurality of preset characterization vectors in a preset space includes:
determining whether a preset space point exists in a preset range of a target space point in the preset space; the target space point is a corresponding space point of the target characterization vector in the preset space; the preset space point is a corresponding space point of the preset characterization vector in the preset space;
determining the target operation track data as real person operation track data under the condition that the preset space point is included in the preset range of the target space point;
and under the condition that the preset space point is not included in the preset range of the target space point, determining that the target operation track data is machine operation track data.
Optionally, the determining the type of the operation trajectory data according to the position relationship between the target characterization vector and a plurality of preset characterization vectors in a preset space includes:
determining whether a target space point in the preset space is in a preset area in the preset space; the target space point is a corresponding space point of the target characterization vector in the preset space; the preset area is as follows: determining a spatial range according to the plurality of preset characterization vectors;
determining the operation track data as real person operation track data under the condition that the target space point is in the preset area;
and under the condition that the target space point is not in the preset area, determining the operation track data as machine operation track data.
Optionally, before preprocessing the target operation trajectory data, the human-computer recognition method further includes:
training the preset human-machine recognition model according to a plurality of groups of sample data;
and each group of the sample data is real person operation track data.
Optionally, the preset human-machine recognition model is a self-encoder model, and the self-encoder model includes: an encoder and a decoder; under the condition that the preset human-machine recognition model is the self-encoder model, training the preset human-machine recognition model according to multiple groups of sample data comprises the following steps:
preprocessing a first group of sample data in the multiple groups of sample data to generate first input sample data;
inputting the first input sample data into an encoder in the preset human-computer recognition model, and obtaining a first characterization vector corresponding to the first input sample data;
inputting the first characterization vector into a decoder in the preset human-computer recognition model to obtain first output data corresponding to the first characterization vector;
determining an error value between the first output data and the first input sample data;
under the condition that the error value is larger than a preset error value, adjusting parameters of the preset human-computer recognition model, and continuing training the preset human-computer recognition model after the parameters are adjusted by using residual sample data until the training of the preset human-computer recognition model is determined to be finished under the condition that the error value between the output data of the decoder and the input sample data of the encoder is smaller than or equal to the preset error value for a first preset number of times; or determining to finish the training of the preset human-computer recognition model under the condition that the training times of the preset human-computer recognition model reach a second preset time.
Optionally, under the condition that the neural network in the preset human-computer recognition model is a recurrent neural network, the preprocessing the target operation trajectory data to generate input data of the preset human-computer recognition model includes:
and carrying out normalization processing on the target operation track data to generate the input data.
Optionally, under the condition that the neural network in the human-computer recognition model is a deep neural network model, the preprocessing the target operation trajectory data to generate input data of a preset human-computer recognition model includes:
acquiring a data sequence corresponding to each preset parameter according to the target operation track data; wherein the preset parameters include: at least one of speed, acceleration and time interval of two adjacent sampling moments when the operation track data is collected;
acquiring target statistics corresponding to each data sequence; wherein the target statistics include: at least one of a maximum value in the data sequence, a minimum value in the data sequence, an average value of the data sequence, and an information entropy of the data sequence;
determining the obtained target statistic as the input data.
The communication bus mentioned in the electronic device may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.
The communication interface is used for communication between the terminal and other equipment.
The Memory may include a Random Access Memory (RAM) or a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the processor.
The Processor may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the device can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.
In yet another embodiment of the present invention, a computer-readable storage medium is further provided, which stores instructions that, when executed on a computer, cause the computer to perform the human-machine identification method described in the above embodiment.
In yet another embodiment, a computer program product containing instructions is provided, which when run on a computer causes the computer to perform the human recognition method described in the above embodiments.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims (12)

1. A man-machine identification method is applied to a server, and is characterized by comprising the following steps:
preprocessing target operation track data to generate input data of a preset human-computer recognition model;
generating a target characterization vector corresponding to the input data through the preset human-computer recognition model;
determining the type of the target operation track data according to the position relation of the target characterization vector and a plurality of preset characterization vectors in a preset space;
wherein the preset characterization vector is: generating a characterization vector by the preset human-computer recognition model according to the real human operation track data; the types of operation trajectory data include: real person operation track data and machine operation track data.
2. The human-computer recognition method according to claim 1, wherein the determining the type of the target operation trajectory data according to the position relationship between the target characterization vector and a plurality of preset characterization vectors in a preset space comprises:
determining whether a preset space point exists in a preset range of a target space point in the preset space; the target space point is a corresponding space point of the target characterization vector in the preset space; the preset space point is a corresponding space point of the preset characterization vector in the preset space;
determining the target operation track data as real person operation track data under the condition that the preset space point is included in the preset range of the target space point;
and under the condition that the preset space point is not included in the preset range of the target space point, determining that the target operation track data is machine operation track data.
3. The human-computer recognition method according to claim 1, wherein the determining the type of the operation trajectory data according to the position relationship between the target characterization vector and a plurality of preset characterization vectors in a preset space comprises:
determining whether a target space point in the preset space is in a preset area in the preset space; the target space point is a corresponding space point of the target characterization vector in the preset space; the preset area is as follows: determining a spatial range according to the plurality of preset characterization vectors;
determining the operation track data as real person operation track data under the condition that the target space point is in the preset area;
and under the condition that the target space point is not in the preset area, determining the operation track data as machine operation track data.
4. The human-machine recognition method of claim 1, wherein before preprocessing the target operation trajectory data, the human-machine recognition method further comprises:
training the preset human-machine recognition model according to a plurality of groups of sample data;
and each group of the sample data is real person operation track data.
5. The human-machine recognition method according to claim 4, wherein the preset human-machine recognition model is a self-encoder model; the self-encoder model includes: an encoder and a decoder;
under the condition that the preset human-machine recognition model is the self-encoder model, training the preset human-machine recognition model according to multiple groups of sample data comprises the following steps:
preprocessing a first group of sample data in the multiple groups of sample data to generate first input sample data;
inputting the first input sample data into an encoder in the preset human-computer recognition model, and obtaining a first characterization vector corresponding to the first input sample data;
inputting the first characterization vector into a decoder in the preset human-computer recognition model to obtain first output data corresponding to the first characterization vector;
determining an error value between the first output data and the first input sample data;
under the condition that the error value is larger than a preset error value, adjusting parameters of the preset human-computer recognition model, and continuing training the preset human-computer recognition model after the parameters are adjusted by using residual sample data until the training of the preset human-computer recognition model is determined to be finished under the condition that the error value between the output data of the decoder and the input sample data of the encoder is smaller than or equal to the preset error value for a first preset number of times; or determining to finish the training of the preset human-computer recognition model under the condition that the training times of the preset human-computer recognition model reach a second preset time.
6. The human-computer recognition method according to claim 1, wherein in a case that the neural network in the preset human-computer recognition model is a recurrent neural network, the preprocessing the target operation trajectory data to generate input data of the preset human-computer recognition model comprises:
and carrying out normalization processing on the target operation track data to generate the input data.
7. The human-computer recognition method according to claim 1, wherein in a case that a neural network in the human-computer recognition model is a deep neural network model, the preprocessing the target operation trajectory data to generate input data of a preset human-computer recognition model includes:
acquiring a data sequence corresponding to each preset parameter according to the target operation track data; wherein the preset parameters include: at least one of speed, acceleration and time interval of two adjacent sampling moments when the operation track data is collected;
acquiring target statistics corresponding to each data sequence; wherein the target statistics include: at least one of a maximum value in the data sequence, a minimum value in the data sequence, an average value of the data sequence, and an information entropy of the data sequence;
determining the obtained target statistic as the input data.
8. A man-machine recognition device applied to a server is characterized by comprising:
the first generation module is used for preprocessing the target operation track data and generating input data of a preset human-computer recognition model;
the second generation module is used for generating a target characterization vector corresponding to the input data through the preset human-computer recognition model;
the type determining module is used for determining the type of the target operation track data according to the position relation of the target characterization vector and a plurality of preset characterization vectors in a preset space;
wherein the preset characterization vector is: generating a characterization vector by the preset human-computer recognition model according to the real human operation track data; the types of operation trajectory data include: real person operation track data and machine operation track data.
9. The human-computer recognition apparatus of claim 8, wherein the type determination module comprises:
the first judgment unit is used for determining whether a preset space point exists in a preset range of a target space point in the preset space; the target space point is a corresponding space point of the target characterization vector in the preset space; the preset space point is a corresponding space point of the preset characterization vector in the preset space;
the first determining unit is used for determining that the target operation track data is real person operation track data under the condition that the preset space point exists in the preset range of the target space point;
and the second determining unit is used for determining the target operation track data as machine operation track data under the condition that the preset space point is not included in the preset range of the target space point.
10. The human-computer recognition apparatus of claim 8, wherein the type determination module comprises:
a second judging unit, configured to determine whether a target space point in the preset space is located in a preset region in the preset space; the target space point is a corresponding space point of the target characterization vector in the preset space; the preset area is as follows: determining a spatial range according to the plurality of preset characterization vectors;
a third determining unit, configured to determine that the operation trajectory data is real-person operation trajectory data when the target space point is within the preset area;
a fourth determining unit, configured to determine that the operation trajectory data is machine operation trajectory data when the target spatial point is not within the preset area.
11. An electronic device, comprising: a processor, a communication interface, a memory, and a communication bus; the processor, the communication interface and the memory complete mutual communication through a communication bus;
a memory for storing a computer program;
a processor for implementing the steps of the human-machine identification method according to any one of claims 1 to 7 when executing the program stored in the memory.
12. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the human-machine recognition method according to any one of claims 1 to 7.
CN202010319326.6A 2020-04-21 2020-04-21 Man-machine recognition method and device, electronic equipment and computer readable storage medium Pending CN111666968A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010319326.6A CN111666968A (en) 2020-04-21 2020-04-21 Man-machine recognition method and device, electronic equipment and computer readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010319326.6A CN111666968A (en) 2020-04-21 2020-04-21 Man-machine recognition method and device, electronic equipment and computer readable storage medium

Publications (1)

Publication Number Publication Date
CN111666968A true CN111666968A (en) 2020-09-15

Family

ID=72382660

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010319326.6A Pending CN111666968A (en) 2020-04-21 2020-04-21 Man-machine recognition method and device, electronic equipment and computer readable storage medium

Country Status (1)

Country Link
CN (1) CN111666968A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115062695A (en) * 2022-06-06 2022-09-16 北京字跳网络技术有限公司 Cheating judgment method, device, equipment and medium

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160315948A1 (en) * 2015-04-21 2016-10-27 Alibaba Group Holding Limited Method and system for identifying a human or machine
US20170323093A1 (en) * 2016-05-05 2017-11-09 Baidu Online Network Technology (Beijing) Co., Ltd. Verification method and apparatus for distinguishing man from machine
CN107679374A (en) * 2017-08-23 2018-02-09 北京三快在线科技有限公司 A kind of man-machine recognition methods and device based on sliding trace, electronic equipment
CN108287989A (en) * 2018-01-18 2018-07-17 北京科技大学 A kind of man-machine recognition methods of sliding identifying code based on track
CN109409049A (en) * 2018-10-10 2019-03-01 北京京东金融科技控股有限公司 The method and apparatus of interactive operation for identification
CN109977651A (en) * 2019-03-14 2019-07-05 广州多益网络股份有限公司 Man-machine recognition methods, device and electronic equipment based on sliding trace
WO2019153604A1 (en) * 2018-02-06 2019-08-15 平安科技（深圳）有限公司 Device and method for creating human/machine identification model, and computer readable storage medium
CN110427737A (en) * 2019-06-20 2019-11-08 平安科技（深圳）有限公司 Man-machine recognition methods, device and the computer equipment of operation behavior
CN110807183A (en) * 2019-10-12 2020-02-18 广州多益网络股份有限公司 Sliding verification code man-machine behavior identification method of multi-dimensional feature system
CN111027643A (en) * 2019-12-26 2020-04-17 北京奇艺世纪科技有限公司 Deep neural network model training method, man-machine interaction recognition method and device, electronic equipment and storage medium

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160315948A1 (en) * 2015-04-21 2016-10-27 Alibaba Group Holding Limited Method and system for identifying a human or machine
US20170323093A1 (en) * 2016-05-05 2017-11-09 Baidu Online Network Technology (Beijing) Co., Ltd. Verification method and apparatus for distinguishing man from machine
CN107679374A (en) * 2017-08-23 2018-02-09 北京三快在线科技有限公司 A kind of man-machine recognition methods and device based on sliding trace, electronic equipment
CN109858211A (en) * 2017-08-23 2019-06-07 北京三快在线科技有限公司 A kind of man-machine recognition methods and device based on sliding trace, electronic equipment
CN108287989A (en) * 2018-01-18 2018-07-17 北京科技大学 A kind of man-machine recognition methods of sliding identifying code based on track
WO2019153604A1 (en) * 2018-02-06 2019-08-15 平安科技（深圳）有限公司 Device and method for creating human/machine identification model, and computer readable storage medium
CN109409049A (en) * 2018-10-10 2019-03-01 北京京东金融科技控股有限公司 The method and apparatus of interactive operation for identification
CN109977651A (en) * 2019-03-14 2019-07-05 广州多益网络股份有限公司 Man-machine recognition methods, device and electronic equipment based on sliding trace
CN110427737A (en) * 2019-06-20 2019-11-08 平安科技（深圳）有限公司 Man-machine recognition methods, device and the computer equipment of operation behavior
CN110807183A (en) * 2019-10-12 2020-02-18 广州多益网络股份有限公司 Sliding verification code man-machine behavior identification method of multi-dimensional feature system
CN111027643A (en) * 2019-12-26 2020-04-17 北京奇艺世纪科技有限公司 Deep neural network model training method, man-machine interaction recognition method and device, electronic equipment and storage medium

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115062695A (en) * 2022-06-06 2022-09-16 北京字跳网络技术有限公司 Cheating judgment method, device, equipment and medium

Similar Documents

Publication Publication Date Title
Xiao et al. An intrusion detection model based on feature reduction and convolutional neural networks
CN111753881B (en) Concept sensitivity-based quantitative recognition defending method against attacks
WO2021189364A1 (en) Method and device for generating adversarial image, equipment, and readable storage medium
CN110084610B (en) Network transaction fraud detection system based on twin neural network
CN109413023B (en) Training of machine recognition model, machine recognition method and device, and electronic equipment
JP6812086B2 (en) Training method for reticulated pattern removal system, reticulated pattern removal method, equipment, equipment and media
JP2022141931A (en) Method and device for training living body detection model, method and apparatus for living body detection, electronic apparatus, storage medium, and computer program
CN111343182B (en) Abnormal flow detection method based on gray level graph
CN110677437A (en) User disguised attack detection method and system based on potential space countermeasure clustering
CN111027643B (en) Training method of deep neural network model, man-machine interaction recognition method, device, electronic equipment and storage medium
CN111079930B (en) Data set quality parameter determining method and device and electronic equipment
CN110929242B (en) Method and system for carrying out attitude-independent continuous user authentication based on wireless signals
CN117495205A (en) Industrial Internet experiment system and method
CN116451594A (en) Training method and device of icing prediction model, prediction method and device and electronic equipment
CN117133277B (en) Virtual character expression control method and system for man-machine interaction
CN111666968A (en) Man-machine recognition method and device, electronic equipment and computer readable storage medium
CN114677556A (en) Countermeasure sample generation method of neural network model and related equipment
JP2023554336A (en) Spatiotemporal deep learning for behavioral biometrics
CN111461177A (en) Image identification method and device
CN114567512B (en) Network intrusion detection method, device and terminal based on improved ART2
CN115393876A (en) Online signature identification method based on neural network
CN115664804A (en) LDoS attack detection method based on radial basis function neural network
CN115314239A (en) Analysis method and related equipment for hidden malicious behaviors based on multi-model fusion
CN111640076B (en) Image complement method and device and electronic equipment
CN114724245A (en) CSI-based incremental learning human body action identification method

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