CN117877098B - Abnormal communication behavior identification method and system based on big data and multidimensional features - Google Patents
Abnormal communication behavior identification method and system based on big data and multidimensional features Download PDFInfo
- Publication number
- CN117877098B CN117877098B CN202410257961.4A CN202410257961A CN117877098B CN 117877098 B CN117877098 B CN 117877098B CN 202410257961 A CN202410257961 A CN 202410257961A CN 117877098 B CN117877098 B CN 117877098B
- Authority
- CN
- China
- Prior art keywords
- counterfeiting
- image
- target
- light
- initial
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 230000006854 communication Effects 0.000 title claims abstract description 34
- 238000004891 communication Methods 0.000 title claims abstract description 26
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 24
- 238000012795 verification Methods 0.000 claims abstract description 77
- 239000000523 sample Substances 0.000 claims description 100
- 238000000605 extraction Methods 0.000 claims description 24
- 238000005286 illumination Methods 0.000 claims description 18
- 239000013074 reference sample Substances 0.000 claims description 15
- 230000001815 facial effect Effects 0.000 claims description 11
- 238000006073 displacement reaction Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000006399 behavior Effects 0.000 description 16
- 238000012549 training Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
Landscapes
- Collating Specific Patterns (AREA)
Abstract
The invention discloses an abnormal communication behavior identification method and system based on big data and multidimensional features, which relate to the technical field of communication protection and comprise the following steps: comprising the following steps: s1, acquiring an initial image; s2, judging the reasonable degree of the image light; s3, generating an anti-counterfeiting light instruction set; s4, acquiring an anti-counterfeiting image; s5, acquiring a target anti-counterfeiting characteristic; s6, establishing an anti-counterfeiting verification algorithm in one-to-one correspondence with the anti-counterfeiting light instruction set; s7, based on the target initial characteristics and the target anti-counterfeiting characteristics, judging whether the anti-counterfeiting verification is met or not by adopting an anti-counterfeiting recognition characteristic algorithm; and S8, if the target anti-counterfeiting verification fails, verification matching is performed based on the target initial characteristic and the target anti-counterfeiting characteristic. The invention has the advantages that: the AI image existing in the image instant messaging process can be effectively identified under the influence of the minimum image instant messaging process, so that the high-precision anti-counterfeiting identification of the AI image is realized, and AI fraud existing in the image instant messaging process can be effectively prevented.
Description
Technical Field
The invention relates to the technical field of communication protection, in particular to an abnormal communication behavior identification method and system based on big data and multidimensional features.
Background
AI face changing refers to changing the face of another person into the face of the person by an AI artificial intelligence technology, and at present, false audio and video such as AI face changing, AI sound changing and the like are utilized to perform fraud and defamation illegal behaviors frequently, so that how to accurately identify and prevent AI face changing images in image instant messaging is indistinct.
In the prior art, anti-fake verification for AI face change is usually stopped on the characteristic distribution difference between image source domain data and target domain data, a large amount of prior data of face information is usually required to be obtained in advance for training, the prior data requirement for the face information is difficult to meet in the image instant messaging process, the conventional anti-fake verification method is difficult to intercept abnormal communication behaviors of 'AI face change' in real time, and based on the scheme, the AI image anti-fake recognition method suitable for the image instant messaging process is provided.
Disclosure of Invention
In order to solve the technical problems, the technical scheme provides the abnormal communication behavior identification method and the abnormal communication behavior identification system based on big data and multidimensional features, and the technical scheme utilizes the AI image anti-counterfeiting identification in the instant messaging process based on the face color change generated by the light source in the image instant messaging process, so that the rapid anti-counterfeiting of the AI image in the image instant messaging process can be realized.
In order to achieve the above purpose, the invention adopts the following technical scheme:
An abnormal communication behavior identification method based on big data and multidimensional features comprises the following steps:
s1, acquiring an anti-counterfeiting image to be identified from a user side, and marking the anti-counterfeiting image as an initial image;
S2, performing target extraction based on the initial image by adopting a target feature extraction algorithm, obtaining target initial features, judging the light rationality of the image based on the target initial features, judging that the target anti-counterfeiting verification passes if the light rationality of the image is judged to pass, and performing step S3 if the light rationality of the image is judged not to pass;
S3, generating an anti-counterfeiting light instruction set;
s4, randomly selecting an anti-counterfeiting light instruction from the anti-counterfeiting light instruction set, sending the anti-counterfeiting light instruction to the user side, acquiring all images of the user side during the anti-counterfeiting light instruction, and marking the images as anti-counterfeiting images;
S5, extracting target features based on the anti-counterfeiting image by adopting a target feature extraction algorithm to obtain target anti-counterfeiting features;
S6, establishing an anti-counterfeiting verification algorithm in one-to-one correspondence with the anti-counterfeiting light instruction set;
s7, based on the initial characteristics of the target and the anti-counterfeiting characteristics of the target, adopting an anti-counterfeiting identification characteristic algorithm to judge whether the anti-counterfeiting verification is met, if so, judging that the anti-counterfeiting verification of the target passes, and if not, judging that the anti-counterfeiting verification of the target fails.
Preferably, the target feature extraction algorithm specifically includes:
Carrying out primary outline recognition on the acquired image, extracting outline feature points of an outline image of at least one object, carrying out image recognition on the outline feature points, judging whether the outline image is a human head outline, if so, judging that the outline image is a human face outline, extracting an image frame with the human face outline, obtaining a plurality of human face outline images, and if not, judging that the human face outline is not present;
And carrying out feature division on the face contour image to obtain face contour region features and background region features, and taking the face contour region features and the background region features as target features.
Preferably, the determining the image light rationality based on the target initial feature specifically includes:
Judging whether at least two face contour images with the face contour region features and the background region features having the displacement exist or not, if so, acquiring at least two face contour images with the face contour region features and the background region features having the displacement, marking the face contour images as light reasonable verification images, and if not, judging that the light reasonable degree of the images does not pass;
Scaling the facial contour region features in the plurality of light reasonable verification images to the same proportion to obtain a plurality of light reasonable verification zoom diagrams;
Grid division is carried out on a plurality of reasonable light verification shrinkage graphs, and at least one grid is randomly selected for verification matching;
Judging whether the grid area luminosity of the verification matching of the plurality of light reasonable verification shrinkage graphs changes or not, if so, judging that the image light rationality passes, and if not, judging that the image light rationality does not pass.
Preferably, the anti-counterfeit verification algorithm for establishing one-to-one correspondence with the anti-counterfeit light instruction set specifically comprises:
acquiring all anti-counterfeiting light characteristics existing in all anti-counterfeiting light instruction sets;
Constructing a face color change feature corresponding to each anti-counterfeiting light feature;
acquiring an instruction time period of each anti-counterfeiting light characteristic of a user side in the execution process of each anti-counterfeiting light instruction;
extracting images corresponding to each anti-counterfeiting light characteristic from the anti-counterfeiting image based on the instruction time period of each anti-counterfeiting light characteristic, and marking the images as light identification images;
judging whether the face color change between the light identification image and the initial image accords with the face color change characteristic corresponding to the anti-counterfeiting light characteristic, if so, judging that the target anti-counterfeiting verification passes; if not, judging that the target anti-counterfeiting verification fails.
Preferably, the construction of the facial color change feature corresponding to each anti-counterfeiting light feature specifically includes:
Setting a plurality of sample illumination environments based on the real environment;
Under different sample illumination environments, image acquisition is carried out on a plurality of reference sample faces to obtain an initial sample image;
Under different sample illumination environments, controlling a user side to output anti-counterfeiting light characteristics, and carrying out image acquisition on a plurality of reference sample faces to obtain characteristic sample images corresponding to the anti-counterfeiting light characteristics;
calculating chromaticity changes of an initial sample image face area and a characteristic sample image face area among the same reference sample faces in the same sample illumination environment;
and (3) recording the chromaticity change of the same reference sample face under all sample illumination environments as the face chromaticity change characteristic corresponding to the anti-counterfeiting light characteristic.
Preferably, the determining whether the anti-counterfeiting verification is met by adopting an anti-counterfeiting recognition feature algorithm based on the target initial feature and the target anti-counterfeiting feature specifically comprises:
Determining an initial sample image closest to the initial target feature, taking the initial sample image as a fitting sample of the initial target feature, and calculating a fitting value between the initial target feature and the fitting sample;
determining a chromaticity change error interval of the target initial feature based on a fitting value between the target initial feature and the fitting sample;
Extracting the facial contour region features of all the light identification images by adopting a target feature extraction algorithm, calculating the average chromaticity of the facial contour region features corresponding to all the light identification images, and marking the average chromaticity as a first anti-counterfeiting chromaticity;
Taking a characteristic sample image of the fitting sample under the anti-counterfeiting light characteristic corresponding to the light identification image, marking the characteristic sample image as the fitting characteristic sample image, calculating the average chromaticity of the face contour area characteristic in the fitting characteristic sample image, and marking the average chromaticity as the second anti-counterfeiting chromaticity;
Calculating the difference between the first anti-counterfeiting chromaticity and the second anti-counterfeiting chromaticity, marking the difference as an anti-counterfeiting chromaticity difference, judging whether the anti-counterfeiting chromaticity difference is within a chromaticity change error interval, if so, judging that the target anti-counterfeiting verification passes, and if not, judging that the target anti-counterfeiting verification fails.
Preferably, the determining the initial sample image closest to the target initial feature as the fitted sample of the target initial feature specifically includes:
calculating average chromaticity of the face contour region features in the target initial features, and marking the average chromaticity as first matching chromaticity;
Calculating average chromaticity of the face contour region features in each initial sample image respectively, and marking the average chromaticity as second matching chromaticity;
Calculating the absolute value of the difference between the first matching chromaticity and the second matching chromaticity to obtain an initial matching value of the target initial characteristic and the initial sample image;
And screening out an initial sample image with the minimum initial matching value as a fitting sample of the target initial characteristic.
Preferably, the calculation formula for calculating the fitting value between the initial feature of the target and the fitting sample is as follows:；
In the method, in the process of the invention, For the fitting value between the initial feature of the target and the fitting sample,/>To fit the second matching chromaticity value corresponding to the sample,/>To fit the initial match value corresponding to the sample.
Preferably, the determining the chromaticity variation error interval of the target initial feature based on the fitting value between the target initial feature and the fitting sample specifically includes:
Taking the chromaticity change of the fitting sample under the anti-counterfeiting light characteristic, and recording the chromaticity change as a chromaticity change reference value ；
The chromaticity change error interval is。
Furthermore, an abnormal communication behavior recognition system based on big data and multidimensional features is provided, which is used for implementing the abnormal communication behavior recognition method based on big data and multidimensional features, and the abnormal communication behavior recognition system comprises the following steps:
the device comprises a basic image feature acquisition module, a first image acquisition module and a second image acquisition module, wherein the basic image feature acquisition module is used for acquiring an anti-counterfeiting image to be identified from a user side and recording the anti-counterfeiting image as an initial image;
The basic feature recognition module is electrically connected with the basic image feature acquisition module and is used for extracting a target based on an initial image by adopting a target feature extraction algorithm to obtain a target initial feature and judging the light rationality of the image based on the target initial feature;
the anti-counterfeiting light control module is used for generating an anti-counterfeiting light instruction set, randomly selecting the anti-counterfeiting light instruction from the anti-counterfeiting light instruction set and sending the anti-counterfeiting light instruction to the user side;
The light image characteristic acquisition module is electrically connected with the anti-counterfeiting light control module and is used for acquiring all images of a user side in an anti-counterfeiting light instruction period and marking the images as anti-counterfeiting images;
The light anti-counterfeiting recognition module is electrically connected with the light image feature acquisition module, and is used for extracting target features based on an anti-counterfeiting image by adopting a target feature extraction algorithm, acquiring target anti-counterfeiting features, establishing an anti-counterfeiting verification algorithm corresponding to an anti-counterfeiting light instruction set one by one, and judging whether the anti-counterfeiting verification algorithm is met or not by adopting the anti-counterfeiting recognition feature algorithm based on the target initial features and the target anti-counterfeiting features
Compared with the prior art, the invention has the beneficial effects that:
The invention provides an abnormal communication behavior recognition scheme based on big data and multidimensional features, adopts a double-section feature verification mode to perform AI image anti-counterfeiting recognition possibly existing in the image instant communication process, firstly performs image light rationality priori judgment of target initial features based on the change of human face shadow features generated when a human face is displaced from a light source in the image instant communication process, does not influence the normal image instant communication process in the recognition stage, and simultaneously performs targeted anti-counterfeiting recognition only on the image instant communication process with problems in light rationality priori judgment, so that AI images existing in the image instant communication process can be effectively recognized, and further high-precision anti-counterfeiting recognition on the AI images is realized under the influence of minimizing the image instant communication process.
Drawings
FIG. 1 is a flow chart of an abnormal communication behavior recognition method based on big data and multidimensional features;
FIG. 2 is a flow chart of a method of the target feature extraction algorithm in the present solution;
FIG. 3 is a flowchart of a method for determining the ray rationality of an image based on initial characteristics of a target in the present solution;
FIG. 4 is a flow chart of a method for establishing an anti-counterfeit authentication algorithm in one-to-one correspondence with an anti-counterfeit light instruction set in the present scheme;
FIG. 5 is a flow chart of a method for constructing a face color change feature corresponding to each anti-counterfeiting light feature in the present solution;
FIG. 6 is a flow chart of a method for determining whether the anti-counterfeit verification is met by adopting an anti-counterfeit identification feature algorithm in the scheme;
fig. 7 is a flow chart of a method for determining a fitted sample of initial features of a target in the present approach.
Detailed Description
The following description is presented to enable one of ordinary skill in the art to make and use the invention. The preferred embodiments in the following description are by way of example only and other obvious variations will occur to those skilled in the art.
Referring to fig. 1, a method for identifying abnormal communication behavior based on big data and multidimensional features includes:
s1, acquiring an anti-counterfeiting image to be identified from a user side, and marking the anti-counterfeiting image as an initial image;
S2, performing target extraction based on the initial image by adopting a target feature extraction algorithm, obtaining target initial features, judging the light rationality of the image based on the target initial features, judging that the target anti-counterfeiting verification passes if the light rationality of the image is judged to pass, and performing step S3 if the light rationality of the image is judged not to pass;
S3, generating an anti-counterfeiting light instruction set;
s4, randomly selecting an anti-counterfeiting light instruction from the anti-counterfeiting light instruction set, sending the anti-counterfeiting light instruction to the user side, acquiring all images of the user side during the anti-counterfeiting light instruction, and marking the images as anti-counterfeiting images;
S5, extracting target features based on the anti-counterfeiting image by adopting a target feature extraction algorithm to obtain target anti-counterfeiting features;
S6, establishing an anti-counterfeiting verification algorithm in one-to-one correspondence with the anti-counterfeiting light instruction set;
s7, based on the initial characteristics of the target and the anti-counterfeiting characteristics of the target, adopting an anti-counterfeiting identification characteristic algorithm to judge whether the anti-counterfeiting verification is met, if so, judging that the anti-counterfeiting verification of the target passes, and if not, judging that the anti-counterfeiting verification of the target fails.
According to the scheme, the AI image anti-counterfeiting possibly existing in the image instant messaging process is carried out in a double-section verification mode, firstly, the image light rationality priori judgment of the target initial characteristic is carried out based on the change of the human face shadow characteristic generated when the human face and the light source are displaced in the image instant messaging process, the normal image instant messaging process is not influenced in the recognition stage, meanwhile, the image instant messaging process with problems in the light rationality priori judgment is subjected to targeted anti-counterfeiting recognition, and the high-precision anti-counterfeiting recognition of the AI image can be effectively realized under the influence of minimizing the image instant messaging process.
Referring to fig. 2, the target feature extraction algorithm specifically includes:
Carrying out primary outline recognition on the acquired image, extracting outline feature points of an outline image of at least one object, carrying out image recognition on the outline feature points, judging whether the outline image is a human head outline, if so, judging that the outline image is a human face outline, extracting an image frame with the human face outline, obtaining a plurality of human face outline images, and if not, judging that the human face outline is not present;
And carrying out feature division on the face contour image to obtain face contour region features and background region features, and taking the face contour region features and the background region features as target features.
Face contour region recognition is a mature prior art in the field, and is not described in detail in this scheme.
Referring to fig. 3, determining the degree of rationality of the image light based on the initial characteristics of the object specifically includes:
Judging whether at least two face contour images with the face contour region features and the background region features having the displacement exist or not, if so, acquiring at least two face contour images with the face contour region features and the background region features having the displacement, marking the face contour images as light reasonable verification images, and if not, judging that the light reasonable degree of the images does not pass;
Scaling the facial contour region features in the plurality of light reasonable verification images to the same proportion to obtain a plurality of light reasonable verification zoom diagrams;
Grid division is carried out on a plurality of reasonable light verification shrinkage graphs, and at least one grid is randomly selected for verification matching;
Judging whether the grid area luminosity of the verification matching of the plurality of light reasonable verification shrinkage graphs changes or not, if so, judging that the image light rationality passes, and if not, judging that the image light rationality does not pass.
It can be understood that, because the light performance of the face area during the AI face-changing is generally the same as the light area of the face area in the training target image, when the AI face-changing is adopted to perform image instant messaging, the face area has low sensitivity to the change of the face shadow feature caused by the change of the ambient light position, and in addition, in the normal communication process, the face inevitably has position change.
Referring to fig. 4, the anti-counterfeit verification algorithm for establishing one-to-one correspondence with the anti-counterfeit light instruction set specifically includes:
acquiring all anti-counterfeiting light characteristics existing in all anti-counterfeiting light instruction sets;
Constructing a face color change feature corresponding to each anti-counterfeiting light feature;
acquiring an instruction time period of each anti-counterfeiting light characteristic of a user side in the execution process of each anti-counterfeiting light instruction;
extracting images corresponding to each anti-counterfeiting light characteristic from the anti-counterfeiting image based on the instruction time period of each anti-counterfeiting light characteristic, and marking the images as light identification images;
judging whether the face color change between the light identification image and the initial image accords with the face color change characteristic corresponding to the anti-counterfeiting light characteristic, if so, judging that the target anti-counterfeiting verification passes; if not, judging that the target anti-counterfeiting verification fails.
Referring to fig. 5, constructing the face color change feature corresponding to each anti-counterfeiting light feature specifically includes:
Setting a plurality of sample illumination environments based on the real environment;
Under different sample illumination environments, image acquisition is carried out on a plurality of reference sample faces to obtain an initial sample image;
Under different sample illumination environments, controlling a user side to output anti-counterfeiting light characteristics, and carrying out image acquisition on a plurality of reference sample faces to obtain characteristic sample images corresponding to the anti-counterfeiting light characteristics;
calculating chromaticity changes of an initial sample image face area and a characteristic sample image face area among the same reference sample faces in the same sample illumination environment;
and (3) recording the chromaticity change of the same reference sample face under all sample illumination environments as the face chromaticity change characteristic corresponding to the anti-counterfeiting light characteristic.
Because in the actual verification process, people with different skin chromaticities have differences on the chromaticity change of the face area generated when the anti-counterfeiting light characteristic changes, people with different chromaticities should be covered as much as possible when the reference sample face is selected based on the differences, and in addition, a plurality of sample illumination environments conforming to the actual environment are constructed based on the actual environment to serve as training samples of the face chromaticity change characteristic under the anti-counterfeiting light characteristic.
Referring to fig. 6, based on the target initial feature and the target anti-counterfeiting feature, determining whether the anti-counterfeiting verification is met by adopting an anti-counterfeiting recognition feature algorithm specifically includes:
Determining an initial sample image closest to the initial target feature, taking the initial sample image as a fitting sample of the initial target feature, and calculating a fitting value between the initial target feature and the fitting sample;
determining a chromaticity change error interval of the target initial feature based on a fitting value between the target initial feature and the fitting sample;
Extracting the facial contour region features of all the light identification images by adopting a target feature extraction algorithm, calculating the average chromaticity of the facial contour region features corresponding to all the light identification images, and marking the average chromaticity as a first anti-counterfeiting chromaticity;
Taking a characteristic sample image of the fitting sample under the anti-counterfeiting light characteristic corresponding to the light identification image, marking the characteristic sample image as the fitting characteristic sample image, calculating the average chromaticity of the face contour area characteristic in the fitting characteristic sample image, and marking the average chromaticity as the second anti-counterfeiting chromaticity;
Calculating the difference between the first anti-counterfeiting chromaticity and the second anti-counterfeiting chromaticity, marking the difference as an anti-counterfeiting chromaticity difference, judging whether the anti-counterfeiting chromaticity difference is within a chromaticity change error interval, if so, judging that the target anti-counterfeiting verification passes, and if not, judging that the target anti-counterfeiting verification fails.
Referring to fig. 7, determining an initial sample image closest to the target initial feature as a fitted sample of the target initial feature specifically includes:
calculating average chromaticity of the face contour region features in the target initial features, and marking the average chromaticity as first matching chromaticity;
Calculating average chromaticity of the face contour region features in each initial sample image respectively, and marking the average chromaticity as second matching chromaticity;
Calculating the absolute value of the difference between the first matching chromaticity and the second matching chromaticity to obtain an initial matching value of the target initial characteristic and the initial sample image;
And screening out an initial sample image with the minimum initial matching value as a fitting sample of the target initial characteristic.
The calculation formula for calculating the fitting value between the initial characteristic of the target and the fitting sample is as follows:；
In the method, in the process of the invention, For the fitting value between the initial feature of the target and the fitting sample,/>To fit the second matching chromaticity value corresponding to the sample,/>To fit the initial match value corresponding to the sample.
Based on the fitting value between the target initial feature and the fitting sample, determining the chromaticity change error interval of the target initial feature specifically includes:
Taking the chromaticity change of the fitting sample under the anti-counterfeiting light characteristic, and recording the chromaticity change as a chromaticity change reference value ；
The chromaticity variation error interval is。
In the scheme, the anti-counterfeiting chromaticity identification is carried out in a sample library priori mode, and all practical conditions are difficult to cover in the sample library, therefore, in the scheme, a fitting calculation mode is adopted to screen the reference sample face and sample illumination environment closest to the sample library, the reference sample face and sample illumination environment are used as fitting samples, a chromaticity change error interval of the target initial characteristic is determined based on fitting values between the fitting samples and the target initial characteristic, the target characteristic with the chromaticity change in the chromaticity change error interval is judged to pass the anti-counterfeiting verification, the false judgment rate of the anti-counterfeiting verification can be effectively reduced, and the anti-counterfeiting identification accuracy in the instant image communication process is improved.
Furthermore, based on the same inventive concept as the above-mentioned abnormal communication behavior recognition method based on big data and multidimensional features, the present solution also provides an abnormal communication behavior recognition system based on big data and multidimensional features, which comprises:
the device comprises a basic image feature acquisition module, a first image acquisition module and a second image acquisition module, wherein the basic image feature acquisition module is used for acquiring an anti-counterfeiting image to be identified from a user side and recording the anti-counterfeiting image as an initial image;
The basic feature recognition module is electrically connected with the basic image feature acquisition module and is used for extracting a target based on an initial image by adopting a target feature extraction algorithm to obtain a target initial feature and judging the light rationality of the image based on the target initial feature;
the anti-counterfeiting light control module is used for generating an anti-counterfeiting light instruction set, randomly selecting the anti-counterfeiting light instruction from the anti-counterfeiting light instruction set and sending the anti-counterfeiting light instruction to the user side;
The light image characteristic acquisition module is electrically connected with the anti-counterfeiting light control module and is used for acquiring all images of a user side in an anti-counterfeiting light instruction period and marking the images as anti-counterfeiting images;
The light anti-counterfeiting recognition module is electrically connected with the light image feature acquisition module, and is used for extracting target features based on anti-counterfeiting images by adopting a target feature extraction algorithm, acquiring target anti-counterfeiting features, establishing an anti-counterfeiting verification algorithm corresponding to an anti-counterfeiting light instruction set one by one, and judging whether the anti-counterfeiting verification is met or not by adopting the anti-counterfeiting recognition feature algorithm based on the target initial features and the target anti-counterfeiting features.
The abnormal communication behavior recognition system based on big data and multidimensional features comprises the following working processes:
step one: the basic image characteristic acquisition module is used for acquiring an anti-counterfeiting image to be identified from a user side and recording the anti-counterfeiting image as an initial image;
Step two: the basic feature recognition module is used for extracting a target based on an initial image by adopting a target feature extraction algorithm, acquiring a target initial feature, judging the light rationality of the image based on the target initial feature, judging that the target anti-counterfeiting verification passes if the light rationality judgment of the image passes, and outputting a signal to the anti-counterfeiting light control module if the light rationality judgment of the image does not pass;
Step three: the anti-counterfeiting light control module is used for generating an anti-counterfeiting light instruction set, randomly selecting an anti-counterfeiting light instruction from the anti-counterfeiting light instruction set and sending the anti-counterfeiting light instruction to the user side;
Step four: the light image characteristic acquisition module is used for acquiring all images of the user side in the anti-counterfeiting light instruction period and marking the images as anti-counterfeiting images;
Step five: the line anti-counterfeiting recognition module is used for extracting target characteristics based on an anti-counterfeiting image by adopting a target characteristic extraction algorithm, obtaining target anti-counterfeiting characteristics, establishing an anti-counterfeiting verification algorithm corresponding to an anti-counterfeiting ray instruction set one by one, judging whether the anti-counterfeiting verification is met or not by adopting the anti-counterfeiting recognition characteristic algorithm based on the target initial characteristics and the target anti-counterfeiting characteristics, if so, judging that the target anti-counterfeiting verification passes, and if not, judging that the target anti-counterfeiting verification fails.
In summary, the invention has the advantages that: the AI image existing in the image instant messaging process can be effectively identified under the influence of the minimum image instant messaging process, so that the high-precision anti-counterfeiting identification of the AI image is realized, and AI fraud existing in the image instant messaging process can be effectively prevented.
The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made therein without departing from the spirit and scope of the invention, which is defined by the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (6)
1. An abnormal communication behavior identification method based on big data and multidimensional features is characterized by comprising the following steps:
s1, acquiring an anti-counterfeiting image to be identified from a user side, and marking the anti-counterfeiting image as an initial image;
S2, performing target extraction based on the initial image by adopting a target feature extraction algorithm, obtaining target initial features, judging the light rationality of the image based on the target initial features, judging that the target anti-counterfeiting verification passes if the light rationality of the image is judged to pass, and performing step S3 if the light rationality of the image is judged not to pass;
S3, generating an anti-counterfeiting light instruction set;
s4, randomly selecting an anti-counterfeiting light instruction from the anti-counterfeiting light instruction set, sending the anti-counterfeiting light instruction to the user side, acquiring all images of the user side during the anti-counterfeiting light instruction, and marking the images as anti-counterfeiting images;
S5, extracting target features based on the anti-counterfeiting image by adopting a target feature extraction algorithm to obtain target anti-counterfeiting features;
S6, establishing an anti-counterfeiting verification algorithm in one-to-one correspondence with the anti-counterfeiting light instruction set;
S7, based on the initial characteristics of the target and the anti-counterfeiting characteristics of the target, adopting an anti-counterfeiting identification characteristic algorithm to judge whether the anti-counterfeiting verification is met, if so, judging that the anti-counterfeiting verification of the target passes, and if not, judging that the anti-counterfeiting verification of the target fails;
the target feature extraction algorithm specifically comprises the following steps:
Carrying out primary outline recognition on the acquired image, extracting outline feature points of an outline image of at least one object, carrying out image recognition on the outline feature points, judging whether the outline image is a human head outline, if so, judging that the outline image is a human face outline, extracting an image frame with the human face outline, obtaining a plurality of human face outline images, and if not, judging that the human face outline is not present;
Carrying out feature division on the face contour image to obtain face contour region features and background region features, and taking the face contour region features and the background region features as target features;
The determining the image light rationality based on the target initial feature specifically comprises the following steps:
Judging whether at least two face contour images with the face contour region features and the background region features having the displacement exist or not, if so, acquiring at least two face contour images with the face contour region features and the background region features having the displacement, marking the face contour images as light reasonable verification images, and if not, judging that the light reasonable degree of the images does not pass;
Scaling the facial contour region features in the plurality of light reasonable verification images to the same proportion to obtain a plurality of light reasonable verification zoom diagrams;
Grid division is carried out on a plurality of reasonable light verification shrinkage graphs, and at least one grid is randomly selected for verification matching;
Judging whether the grid area luminosity of the verification matching of the plurality of light reasonable verification shrinkage graphs changes or not, if so, judging that the image light rationality passes, and if not, judging that the image light rationality does not pass;
the anti-counterfeiting verification algorithm for establishing one-to-one correspondence with the anti-counterfeiting light instruction set specifically comprises the following steps:
acquiring all anti-counterfeiting light characteristics existing in all anti-counterfeiting light instruction sets;
Constructing a face color change feature corresponding to each anti-counterfeiting light feature;
acquiring an instruction time period of each anti-counterfeiting light characteristic of a user side in the execution process of each anti-counterfeiting light instruction;
extracting images corresponding to each anti-counterfeiting light characteristic from the anti-counterfeiting image based on the instruction time period of each anti-counterfeiting light characteristic, and marking the images as light identification images;
judging whether the face color change between the light identification image and the initial image accords with the face color change characteristic corresponding to the anti-counterfeiting light characteristic, if so, judging that the target anti-counterfeiting verification passes; if not, judging that the target anti-counterfeiting verification fails;
The method for judging whether the anti-counterfeiting verification is met by adopting an anti-counterfeiting recognition feature algorithm based on the target initial feature and the target anti-counterfeiting feature specifically comprises the following steps:
Determining an initial sample image closest to the initial target feature, taking the initial sample image as a fitting sample of the initial target feature, and calculating a fitting value between the initial target feature and the fitting sample;
determining a chromaticity change error interval of the target initial feature based on a fitting value between the target initial feature and the fitting sample;
Extracting the facial contour region features of all the light identification images by adopting a target feature extraction algorithm, calculating the average chromaticity of the facial contour region features corresponding to all the light identification images, and marking the average chromaticity as a first anti-counterfeiting chromaticity;
Taking a characteristic sample image of the fitting sample under the anti-counterfeiting light characteristic corresponding to the light identification image, marking the characteristic sample image as the fitting characteristic sample image, calculating the average chromaticity of the face contour area characteristic in the fitting characteristic sample image, and marking the average chromaticity as the second anti-counterfeiting chromaticity;
Calculating the difference between the first anti-counterfeiting chromaticity and the second anti-counterfeiting chromaticity, marking the difference as an anti-counterfeiting chromaticity difference, judging whether the anti-counterfeiting chromaticity difference is within a chromaticity change error interval, if so, judging that the target anti-counterfeiting verification passes, and if not, judging that the target anti-counterfeiting verification fails.
2. The abnormal communication behavior recognition method based on big data and multidimensional features according to claim 1, wherein the constructing the facial color change feature corresponding to each anti-fake light feature specifically comprises:
Setting a plurality of sample illumination environments based on the real environment;
Under different sample illumination environments, image acquisition is carried out on a plurality of reference sample faces to obtain an initial sample image;
Under different sample illumination environments, controlling a user side to output anti-counterfeiting light characteristics, and carrying out image acquisition on a plurality of reference sample faces to obtain characteristic sample images corresponding to the anti-counterfeiting light characteristics;
calculating chromaticity changes of an initial sample image face area and a characteristic sample image face area among the same reference sample faces in the same sample illumination environment;
and (3) recording the chromaticity change of the same reference sample face under all sample illumination environments as the face chromaticity change characteristic corresponding to the anti-counterfeiting light characteristic.
3. The abnormal communication behavior recognition method based on big data and multidimensional features according to claim 2, wherein the determining the initial sample image closest to the target initial feature as the fitted sample of the target initial feature specifically comprises:
calculating average chromaticity of the face contour region features in the target initial features, and marking the average chromaticity as first matching chromaticity;
Calculating average chromaticity of the face contour region features in each initial sample image respectively, and marking the average chromaticity as second matching chromaticity;
Calculating the absolute value of the difference between the first matching chromaticity and the second matching chromaticity to obtain an initial matching value of the target initial characteristic and the initial sample image;
And screening out an initial sample image with the minimum initial matching value as a fitting sample of the target initial characteristic.
4. The abnormal communication behavior recognition method based on big data and multidimensional features according to claim 3, wherein the calculation formula for calculating the fitting value between the initial feature of the target and the fitting sample is as follows:；
In the method, in the process of the invention, For the fitting value between the initial feature of the target and the fitting sample,/>To fit the second matching chromaticity value corresponding to the sample,/>To fit the initial match value corresponding to the sample.
5. The abnormal communication behavior recognition method based on big data and multidimensional features according to claim 4, wherein the determining the chromaticity variation error interval of the target initial feature based on the fitting value between the target initial feature and the fitting sample specifically comprises:
Taking the chromaticity change of the fitting sample under the anti-counterfeiting light characteristic, and recording the chromaticity change as a chromaticity change reference value ；
The chromaticity change error interval is。
6. An abnormal communication behavior recognition system based on big data and multidimensional features, which is used for realizing the abnormal communication behavior recognition method based on big data and multidimensional features as claimed in any one of claims 1-5, comprising:
the device comprises a basic image feature acquisition module, a first image acquisition module and a second image acquisition module, wherein the basic image feature acquisition module is used for acquiring an anti-counterfeiting image to be identified from a user side and recording the anti-counterfeiting image as an initial image;
The basic feature recognition module is electrically connected with the basic image feature acquisition module and is used for extracting a target based on an initial image by adopting a target feature extraction algorithm to obtain a target initial feature and judging the light rationality of the image based on the target initial feature;
the anti-counterfeiting light control module is used for generating an anti-counterfeiting light instruction set, randomly selecting the anti-counterfeiting light instruction from the anti-counterfeiting light instruction set and sending the anti-counterfeiting light instruction to the user side;
The light image characteristic acquisition module is electrically connected with the anti-counterfeiting light control module and is used for acquiring all images of a user side in an anti-counterfeiting light instruction period and marking the images as anti-counterfeiting images;
The light anti-counterfeiting recognition module is electrically connected with the light image feature acquisition module, and is used for extracting target features based on anti-counterfeiting images by adopting a target feature extraction algorithm, acquiring target anti-counterfeiting features, establishing an anti-counterfeiting verification algorithm corresponding to an anti-counterfeiting light instruction set one by one, and judging whether the anti-counterfeiting verification is met or not by adopting the anti-counterfeiting recognition feature algorithm based on the target initial features and the target anti-counterfeiting features.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410257961.4A CN117877098B (en) | 2024-03-07 | 2024-03-07 | Abnormal communication behavior identification method and system based on big data and multidimensional features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410257961.4A CN117877098B (en) | 2024-03-07 | 2024-03-07 | Abnormal communication behavior identification method and system based on big data and multidimensional features |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117877098A CN117877098A (en) | 2024-04-12 |
| CN117877098B true CN117877098B (en) | 2024-05-28 |
Family
ID=90579581
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410257961.4A Active CN117877098B (en) | 2024-03-07 | 2024-03-07 | Abnormal communication behavior identification method and system based on big data and multidimensional features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117877098B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20090032295A (en) * | 2007-09-27 | 2009-04-01 | 삼성테크윈 주식회사 | Apparatus and method for detecting face |
| WO2013131407A1 (en) * | 2012-03-08 | 2013-09-12 | 无锡中科奥森科技有限公司 | Double verification face anti-counterfeiting method and device |
| WO2016132113A1 (en) * | 2015-02-16 | 2016-08-25 | University Of Surrey | Three dimensional modelling |
| CN109325437A (en) * | 2018-09-17 | 2019-02-12 | 北京旷视科技有限公司 | Image processing method, device and system |
| CN112560989A (en) * | 2020-12-25 | 2021-03-26 | 烟台职业学院 | Artificial intelligence anti-counterfeiting image identification method and system based on big data |
| CN116798130A (en) * | 2023-06-15 | 2023-09-22 | 广州朗国电子科技股份有限公司 | Face anti-counterfeiting method, device and storage medium |
| CN117176858A (en) * | 2023-11-03 | 2023-12-05 | 深圳市图麟科技有限公司 | Communication terminal information acquisition method and system |
-
2024
- 2024-03-07 CN CN202410257961.4A patent/CN117877098B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20090032295A (en) * | 2007-09-27 | 2009-04-01 | 삼성테크윈 주식회사 | Apparatus and method for detecting face |
| WO2013131407A1 (en) * | 2012-03-08 | 2013-09-12 | 无锡中科奥森科技有限公司 | Double verification face anti-counterfeiting method and device |
| WO2016132113A1 (en) * | 2015-02-16 | 2016-08-25 | University Of Surrey | Three dimensional modelling |
| CN109325437A (en) * | 2018-09-17 | 2019-02-12 | 北京旷视科技有限公司 | Image processing method, device and system |
| CN112560989A (en) * | 2020-12-25 | 2021-03-26 | 烟台职业学院 | Artificial intelligence anti-counterfeiting image identification method and system based on big data |
| CN116798130A (en) * | 2023-06-15 | 2023-09-22 | 广州朗国电子科技股份有限公司 | Face anti-counterfeiting method, device and storage medium |
| CN117176858A (en) * | 2023-11-03 | 2023-12-05 | 深圳市图麟科技有限公司 | Communication terminal information acquisition method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117877098A (en) | 2024-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8098904B2 (en) | Automatic face detection and identity masking in images, and applications thereof | |
| EP1255225A2 (en) | Method for detecting eye and mouth positions in a digital image | |
| CN107239735A (en) | A kind of biopsy method and system based on video analysis | |
| CN112801057B (en) | Image processing method, image processing device, computer equipment and storage medium | |
| JP2000003452A (en) | Method for detecting face surface in digital picture, its detecting device, picture judging method, picture judging device and computer readable record medium | |
| CN112818722B (en) | Modular dynamic configurable living body face recognition system | |
| Christlein et al. | A study on features for the detection of copy-move forgeries | |
| WO2004055715A1 (en) | Expression invariant face recognition | |
| CN113312965B (en) | Face unknown spoofing attack living body detection method and system | |
| JP3490910B2 (en) | Face area detection device | |
| CN110059607B (en) | Living body multiplex detection method, living body multiplex detection device, computer equipment and storage medium | |
| CN113793336A (en) | Method, device and equipment for detecting blood cells and readable storage medium | |
| CN112329851A (en) | Icon detection method and device and computer readable storage medium | |
| WO2023165616A1 (en) | Method and system for detecting concealed backdoor of image model, storage medium, and terminal | |
| Nguyen et al. | Face presentation attack detection based on a statistical model of image noise | |
| JP2000331167A (en) | Method and device for collating facial image | |
| CN112200075B (en) | Human face anti-counterfeiting method based on anomaly detection | |
| CN117877098B (en) | Abnormal communication behavior identification method and system based on big data and multidimensional features | |
| CN113468954A (en) | Face counterfeiting detection method based on local area features under multiple channels | |
| JP3962517B2 (en) | Face detection method and apparatus, and computer-readable medium | |
| CN112699811B (en) | Living body detection method, living body detection device, living body detection apparatus, living body detection storage medium, and program product | |
| CN111353348B (en) | Image processing method, device, acquisition equipment and storage medium | |
| CN116152932A (en) | Living body detection method and related equipment | |
| CN111241919A (en) | Face living body detection method and system combining optical verification code | |
| CN112560705A (en) | Face detection method and device and electronic equipment |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |