Image determination device, image determination method, and image determination program

Abstract

To provide an image determination device, an image determination method, and an image determination program that enable an AI image to be detected with high accuracy.SOLUTION: An image determination device 1 includes: an image collection unit 11 that collects face images generated automatically by generative AI; an authentication processing unit 12 that performs predetermined face authentication processing for each of the collected face images and counts the number of other face images determined as faces identical to each of the face images; a master registration unit 13 that stores, as a masterpiece face, a face image counted by a predetermined number of times or more, among the collected face images; a determination unit 14 that determines whether an object image is a face identical to each of masterpiece faces through face authentication; and an output unit 15 that when there is a masterpiece face determined as an identical face, outputs a fact that the object image is an AI image as a determination result.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a technology for detecting images generated by AI. [Background technology]

[0002] In recent years, there has been rapid progress in the technology known as generative AI, which automatically generates images from specified descriptions. For this reason, it has been pointed out that, for example, face images generated by generative AI may lead to fraudulent activities in online identity verification, such as impersonation or registration of a persona that does not actually exist. In response to this, a system has been developed for discriminating between automatically generated images (AI images) and real photographic images (see, for example, Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] AI image detector, Internet<https: / / huggingface.co / spaces / umm-maybe / AI-image-detector> Summary of the Invention [Problem to be solved by the invention]

[0004] However, conventional systems for detecting AI images use AI that has learned AI images and non-AI images, or use features of the AI ​​images (such as blurring of the entire image) for judgment, which tend to result in erroneous judgments. Therefore, further improvement in the judgment accuracy is desired.

[0005] The present invention aims to provide an image judgment device, an image judgment method, and an image judgment program that can detect AI images with high accuracy. [Means for solving the problem]

[0006] The image judgment device of the present invention comprises an image collection unit that collects facial images automatically generated by a generation AI, an authentication processing unit that performs a predetermined face recognition process on each of the facial images collected by the image collection unit and counts the number of other facial images that are judged to be the same face as each facial image, a master registration unit that stores, among the facial images collected by the image collection unit, facial images whose number counted by the authentication processing unit is a predetermined number or more as Maspi faces as a master registration unit, a judgment unit that judges whether a target image is the same face as each of the Maspi faces by facial recognition, and an output unit that outputs as a judgment result that the target image is an AI image when there is a Maspi face that is judged to be the same face by the judgment unit.

[0007] The output unit may output, as the determination result, information indicating which of the Maspi faces the face has been determined to be the same as.

[0008] The image collection unit may collect face images generated by a plurality of types of generation AI.

[0009] The image judgment method of the present invention includes an image collection step of collecting facial images automatically generated by a generation AI, an authentication processing step of performing a predetermined face recognition process on each of the facial images collected in the image collection step and counting the number of other facial images judged to be the same face as each facial image, a master registration step of storing as a master face those facial images collected in the image collection step whose number counted by the authentication processing step is a predetermined number or more, a judgment step of judging whether a target image is the same face as each of the master faces by facial recognition, and an output step of outputting as a judgment result that the target image is an AI image if there is a master face judged to be the same face in the judgment step.

[0010] An image assessment program according to the present invention causes a computer to function as the image assessment device. Effect of the Invention

[0011] According to the present invention, AI images can be detected with high accuracy. [Brief description of the drawings]

[0012] [Figure 1] 1 is a block diagram showing a functional configuration of an image assessment device according to an embodiment. FIG. [Diagram 2] 4 is a flowchart showing a flow of processing by the image assessment device in the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] An example of an embodiment of the present invention will now be described. The image assessment device of this embodiment stores a large number of facial images extracted by facial recognition from facial images generated by various generation AIs as AI images, and uses facial recognition to determine whether the facial image to be assessed is an AI image.

[0014] FIG. 1 is a block diagram showing a functional configuration of an image assessment device 1 according to this embodiment. The image assessment device 1 is an information processing device including a control unit 10, a storage unit 20, and various input / output interfaces.

[0015] The control unit 10 is a part that controls the entire image assessment device 1, and realizes each function in this embodiment by appropriately reading and executing various programs stored in the storage unit 20. The control unit 10 may be a CPU. The storage unit 20 is a storage area for various programs for causing the hardware group to function as the image assessment device 1, various data, and the like, and may be a ROM, a RAM, a flash memory, a hard disk drive (HDD), or the like.

[0016] The control unit 10 executes software (image determination program) stored in the storage unit 20 to function as an image collection unit 11, an authentication processing unit 12, a master registration unit 13, a determination unit 14, and an output unit 15.

[0017] The image collection unit 11 collects face images that are automatically generated by a generation AI. These face images may be generated by the image assessment device 1 itself, or may be input from the outside. Furthermore, the image collection unit 11 may collect various AI images by collecting face images generated by a plurality of types of generation AI.

[0018] The authentication processing unit 12 performs a predetermined face authentication process on each of the face images collected by the image collection unit 11 in order to compare them with other face images, and counts the number of other face images that have been determined to be the same face as each face image.

[0019] The master registration unit 13 stores in the storage unit 20, as a master piece face, face images of which the number counted by the authentication processing unit 12 is equal to or greater than a predetermined number, among the face images collected by the image collection unit 11. It should be noted that, among the face images determined to be the same face, only one of them or a predetermined number of them (the most significant ones counted) may be stored.

[0020] The determination unit 14 determines whether or not the input target image is the same face as each of the Maspi faces by face recognition.

[0021] When there is a Maspi face that has been determined to be the same face by the determination unit 14, the output unit 15 outputs, as a determination result, that the target image is an AI image. Furthermore, the output unit 15 may also output, as a determination result, information on which of the Maspi faces the face has been determined to be the same as.

[0022] FIG. 2 is a flowchart showing the flow of processing by the image assessment device 1 in this embodiment. In step S1, the image collection unit 11 collects a large number (n pieces) of face images generated by a generation AI or generates them by itself, which are denoted as M1 to Mn. In step S2, the authentication processing unit 12 performs a predetermined face authentication process for each of M1 to Mn, and counts the number of images determined to be the same face as Mk (1≦k≦n). In step S3, the master registration unit 13 regards the face images (m pieces) in which the number of identical faces counted in step S2 is greater than or equal to a certain number as master faces, and stores them as MP1 to MPm.

[0023] In step S4, the determination unit 14 performs face recognition for each of MP1 to MPm for a target image that needs to be determined as being an AI image or not, and searches for an identical face. In step S5, the output unit 15 determines whether or not there is a Maspi face that has been recognized as the same face in step S4. If the determination is YES, the process proceeds to step S6, and if the determination is NO, the process proceeds to step S7. In step S6, the output unit 15 determines that the target image is an AI image automatically generated by a generation AI, and outputs the determination result. In step S7, the output unit 15 determines that the target image is a non-AI image, and outputs the determination result.

[0024] According to this embodiment, the image judgment device 1 collects facial images automatically generated by a generation AI and registers a Maspi face that has a large number of other facial images that are judged to be the same face through facial recognition, and when a target image is judged to be the same as a Maspi face through facial recognition, it can detect that the target image is an AI image. In other words, since it is difficult to input conditions for the generation AI to generate a different person, the image judgment device 1 utilizes the AI's tendency to mass-generate similar facial images called Maspi faces, and by appropriately registering these Maspi faces, it is possible to accurately detect through facial recognition that the target image is an AI image. As a result, it is possible to prevent with a high degree of accuracy, for example, attacks that involve the use of facial images easily generated by AI to register fake personalities.

[0025] In addition, the image judgment device 1 can obtain a Maspi face, which is a face that is likely to be an AI image, and when it judges a target image to be an AI image, it can clearly indicate which Maspi face it resembles and that images of that face are generated in large quantities. Therefore, the image judgment device 1 can show the user the reason why the target image was judged to be an AI image, and a highly reliable system can be constructed.

[0026] Furthermore, the image judgment device 1 can collect face images generated by multiple types of generation AI, thereby comprehensively registering mask faces and improving the detection accuracy of AI images.

[0027] Furthermore, this will make it possible to curb fraudulent activities using AI images, for example, and thereby contribute to Goal 9 of the United Nations-led Sustainable Development Goals (SDGs), which is to "build resilient infrastructure, promote sustainable industrialization and foster innovation."

[0028] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments. Furthermore, the effects described in the above-described embodiments are merely a list of the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the embodiments.

[0029] The image assessment device 1 may use other assessment techniques in combination with the image assessment method of this embodiment. For example, images created using AI that learns from real people and generates similar faces are likely to be identified as the person in question using facial recognition, but such targeted AI images that recreate the faces of specific people can be detected by combining them with other technologies.

[0030] The image judgment method by the image judgment device 1 is realized by software. When realized by software, a program constituting this software is installed in an information processing device (computer). Moreover, these programs may be recorded on a removable medium such as a CD-ROM and distributed to users, or may be distributed by being downloaded to the user's computer via a network. Furthermore, these programs may be provided to the user's computer as a Web service via a network without being downloaded. [Explanation of symbols]

[0031] 1. Image assessment device 10 Control section 11 Image Acquisition Section 12 Authentication processing section 13 Master Registration Section 14 Judgment section 15 Output section 20 Memory section

Claims

1. An image collection unit that collects face images automatically generated by the generation AI; an authentication processing unit that performs a predetermined face authentication process on each of the face images collected by the image collecting unit and counts the number of other face images that are determined to be the same face as each face image; a master registration unit that stores, as a master face, face images among the face images collected by the image collection unit, the number of which is counted by the authentication processing unit being equal to or greater than a predetermined number; A determination unit that determines whether a target image is a face identical to each of the Maspi faces by face recognition; An image judgment device comprising: an output unit that outputs as a judgment result that the target image is an AI image when a Maspi face is judged to be the same face by the judgment unit.

2. The image judgment device according to claim 1 , wherein the output unit outputs information indicating which of the Maspi faces the face is judged to be the same as the judgment result.

3. The image assessment device according to claim 1 or claim 2, wherein the image collection unit collects face images generated by a plurality of types of generation AI.

4. An image collection step of collecting face images automatically generated by the generation AI; an authentication processing step of performing a predetermined face authentication process on each of the face images collected in the image collection step, and counting the number of other face images determined to be the same face as each face image; a master registration step of storing, as a master face, face images among the face images collected in the image collection step, the number of which is equal to or greater than a predetermined number counted in the authentication processing step; a determination step of determining whether a target image is a face identical to each of the Maspi faces by face authentication; An image judgment method executed by a computer, comprising: an output step of outputting, as a judgment result, that the target image is an AI image if there is a Maspi face that is judged to be the same face in the judgment step.

5. An image judgment program for causing a computer to function as the image judgment device according to claim 1 or 2.

Images