Method, device and equipment for identifying counterfeit making information, medium and product

Abstract

The invention discloses a method, a device, equipment, a medium and a product for identifying fake information, relates to the technical field of image recognition, can be applied to the field of financial science and technology, and comprises the following steps: determining target feature data of a video to be identified, the target feature data being continuous frame data, the frame data comprising face image information; processing the continuous frame data by using a pre-trained information identification model to obtain long-time sequence pseudo-making information of the video to be identified, the long-time sequence pseudo-making information being used for indicating the difference between the continuous frame data; and processing the long-time sequence counterfeiting information to obtain an authenticity identification result of the to-be-identified video. According to the method, the counterfeit making conclusion is determined according to the difference of each frame of data, the method is not limited to some feature information, the counterfeit making information with records and the counterfeit making information without records can be identified, and whether the face image in the video is the counterfeit making information or not can be accurately, comprehensively and efficiently identified.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, which can be applied to the field of financial technology, and in particular to a method, device, equipment, medium and product for identifying counterfeit information. Background Art

[0002] Deep fakes are fake face, portrait, audio and video content created based on deep learning technology, such as facial image information changes, expression manipulation, lip manipulation, voice conversion, etc. As one of the areas that carries the most identity information, the face greatly affects people's judgment of the authenticity of information content. Improper use will make fake face technology a powerful tool for spreading rumors, slander, and attacks, which may have a destructive impact on personal reputation, the credibility of online information, and even social stability and national security.

[0003] At present, the methods for identifying forged information include feature detection and model detection. Feature detection is to determine whether there is forged information in the face, portrait, audio and video content by identifying necessary feature information. Necessary feature information includes blinking movements, pulse signals, eye reflection direction, smoothness of facial lines, etc. Model detection is to input the audio and video content to be identified into a trained model, and determine whether there is forged information in the audio and video content based on the model output results.

[0004] However, the versatility and detection efficiency of the feature detection method are low. The model in the model detection method is trained with certain specific forged faces as training samples. It can identify known forged information well, but it is difficult to identify unknown forged information. Summary of the invention

[0005] The present invention provides a method, device, equipment, medium and product for identifying forged information, which can be applied to the field of financial technology and aims to accurately, comprehensively and efficiently identify whether a facial image in a video is forged information.

[0006] According to one aspect of the present invention, there is provided a method for identifying forged information, the method comprising:

[0007] Determine target feature data of the video to be identified, wherein the target feature data is continuous frame data, and the frame data includes face image information;

[0008] Using a pre-trained information identification model, continuous frame data is processed to obtain long-term forgery information of the video to be identified, wherein the long-term forgery information is used to indicate the difference between continuous frame data;

[0009] The long-term counterfeit information is processed to obtain the authenticity identification result of the video to be identified.

[0010] According to another aspect of the present invention, a device for identifying forged information is provided. The device for identifying forged information is used to implement the method for identifying forged information in any embodiment of the present invention. The device comprises:

[0011] An acquisition module is used to determine target feature data of the video to be identified, wherein the target feature data is continuous frame data, and the frame data includes face image information;

[0012] A determination module, used to process continuous frame data using a pre-trained information identification model to obtain long-term forgery information of the video to be identified, wherein the long-term forgery information is used to indicate the difference between continuous frame data;

[0013] The identification module is used to process the long-time sequence forged information to obtain the authenticity identification result of the video to be identified.

[0014] According to another aspect of the present invention, there is provided an electronic device, the electronic device comprising:

[0015] at least one processor; and a memory communicatively coupled to the at least one processor;

[0016] The memory stores a computer program executable by at least one processor, and the computer program is executed by at least one processor so that the at least one processor can execute the method for identifying counterfeit information in any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium stores computer instructions. The computer instructions are used to enable a processor to implement the method for identifying forged information in any embodiment of the present invention when executed.

[0018] According to another aspect of the present invention, a computer program product is provided. The computer program product includes a computer program. When the computer program is executed by a processor, the method for identifying counterfeit information according to any embodiment of the present invention is implemented.

[0019] The method for identifying forged information of the present invention comprises: determining target feature data of a video to be identified, the target feature data being continuous frame data, the frame data including facial image information; using a pre-trained information identification model to process the continuous frame data to obtain long-term forged information of the video to be identified, the long-term forged information being used to indicate the difference between continuous frame data; processing the long-term forged information to obtain an authenticity identification result of the video to be identified. The present invention will analyze the video to be identified, obtain continuous frame data after aligning the face, and then identify the difference features between the continuous frame data (i.e., long-term forged information), and determine whether the video to be identified has forged information (i.e., authenticity identification result) based on the long-term forged information. The forgery conclusion is determined based on the difference of each frame data, which is not limited to certain feature information. Both forged information with and without records can be identified. The identification method has high versatility and accuracy, and does not need to determine feature information such as blinking movements, pulse signals, eye reflection direction, and smoothness of facial lines. The detection efficiency is high, that is, it can accurately, comprehensively, and efficiently identify whether the face image in the video is forged information. The low versatility and detection efficiency of the feature detection method are solved. The model in the model detection method is trained with certain specific forged faces as training samples, which can well identify known forged information, but is difficult to identify unknown forged information.

[0020] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0022] Figure 1 It is a flow chart of a method for identifying forged information provided by the present invention;

[0023] Figure 2 It is a flow chart of another method for identifying forged information provided by the present invention;

[0024] Figure 3 It is a structural schematic diagram of a device for identifying counterfeit information provided by the present invention;

[0025] Figure 4 It is a structural schematic diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", "initial", "intermediate", "candidate", "alternative", "target", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] The information collected by the present invention is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0029] The rapid development of deep fake technology has given rise to major challenges to the authenticity and security of video content. Deep fake videos can imitate the performance of real people in videos with high precision, and traditional face recognition technology is often unable to identify the authenticity of deep fake videos. This problem not only threatens personal privacy and security, but may also cause a crisis of social trust. However, challenges also breed opportunities. With the development of computer vision, especially the continuous advancement of deep learning technology, it is possible to effectively identify and prevent deep fakes. Based on the above background, the present invention explores an innovative video deep fake face identification technology, aiming to provide strong technical support for network security and personal information protection. Deep fake is a fake video created using deep learning technology. For example, facial information is replaced or modified, so that the target person appears to be saying or doing things that have never happened in fact, which will bring serious threats to social media, network security and the legal field.

[0030] Figure 1 This is a flow chart of a method for identifying forged information provided by the present invention. This embodiment can be applied to improving the efficiency and accuracy of identifying forged information, improving the universality of the forged identification method, etc. The method can be executed by the device for identifying forged information provided by the present invention. The device can be implemented in the form of hardware and / or software. In a specific embodiment, the device can be integrated in an electronic device. The following embodiments will be described by taking the device integrated in an electronic device as an example. Figure 1 , the method specifically comprises the following steps:

[0031] S101, determining target feature data of a video to be identified.

[0032] The video to be identified can be understood as a video that needs to be authenticated. The authenticity authentication of the present invention refers to determining whether the facial image information in the video is forged information. Video is essentially composed of a series of continuous static images (i.e., frames). The dynamic visual effect created by these images being played continuously at a certain rate is the video. The target feature data is continuous frame data, and each frame data includes facial image information. The purpose of the present invention is to identify whether the facial state image in the aforementioned facial image information is information of synthesis, splicing, or change of partial features.

[0033] In one embodiment, S101 may specifically include: after receiving the video to be identified, using a pre-set face recognition model to process the video to be identified to obtain at least one frame data and at least one frame data of face image information; using a pre-set face alignment model to process at least one frame data of face image information to obtain continuous frame data.

[0034] Among them, the pre-set face recognition model can be understood as a lightweight face detection model, which is used to extract the face image and frame data containing the face image in the video to be identified, and can also be understood as parsing each video frame and extracting the face image in each video frame. The face image information in the extracted frame data may be in different positions in the image, and if the connection between the frame data is not smooth, it may cause the problem of misjudgment. Therefore, after obtaining at least one frame data and at least one frame data of face image information, the present invention will align the face of the same person in different frame data with the help of a pre-set face alignment model, so as to subsequently determine whether the face in the video is forged according to the difference in the frame data.

[0035] It is worth noting that no matter how many faces there are in the video, the present invention will use a face-by-face analysis method to determine whether the facial information is forged. For example, when there are two faces in the video, the two face images are parsed into two continuous frame data respectively, and then processed in parallel to ensure the identification accuracy of each face image. Therefore, the continuous frame data in the present invention is a video sequence of a single face image.

[0036] S102: Using a pre-trained information identification model, continuous frame data is processed to obtain long-term forged information of the video to be identified.

[0037] Among them, the pre-trained information identification model can be understood as an algorithm trained in advance for determining abnormal feature information between multiple frame data of the video to be identified, and the long-term counterfeit information is used to indicate the differences between consecutive frame data, for example, the facial images between two or more consecutive frame data are not connected, some feature information in the facial images between two or more consecutive frame data is inconsistent (for example, eye size, eye direction, etc.), and some feature information in the interval of two or more frame data is inconsistent, etc., which is used to evaluate whether the facial information in the video to be identified is forged information, and the present invention is not limited to this.

[0038] The information identification model includes a target neural network model and a target structure model. The target neural network model can be understood as a pre-trained three-dimensional convolutional neural network (3D-CNN). It is a deep learning model extended from a two-dimensional convolutional neural network (2D-CNN) and can process data with a three-dimensional structure, such as continuous image frames in a video. Compared with 2D-CNN, 3D-CNN not only considers the two-dimensional spatial information of the image, but also considers the changes in the time dimension. It is more powerful in capturing spatiotemporal features and can better understand the dynamics of human faces in videos. Specifically, 3D-CNN has a unique three-dimensional convolution kernel that can capture spatiotemporal features in continuous frame data such as videos. 3D-CNN can simultaneously process the two-dimensional spatial information and time series information of images. This feature makes it show significant advantages in analyzing the dynamics of human faces, expression changes, and posture transfers in videos. Through 3D convolution, the model can learn the three-dimensional morphology and motion patterns of human faces from continuous image frames, which is crucial for the detection of deep fake videos, because deep fakes often leave subtle traces in dynamic changes and coherence.

[0039] The target structure model can be understood as Transformer (a lightweight deep learning architecture) for natural language processing tasks. By calculating the dependencies between elements at all positions in the sequence, it can efficiently capture long-distance dependencies in sequence modeling. In video deepfake detection, Transformer can identify complex patterns of facial dynamics in video sequences, which has significant advantages in detecting deepfake videos. Specifically, the self-attention mechanism of the Transformer structure can effectively process long sequence inputs and capture the dependencies between elements in the sequence, no matter how far apart these elements are. Through the self-attention mechanism, Transformer can focus on the global changes in facial dynamics in video sequences and identify patterns and anomalies related to deepfakes, even if these patterns may span multiple frames. This precise capture of long-distance dependencies, combined with the spatial perception capabilities of 3D-CNN, can bring new breakthroughs to deepfake detection technology.

[0040] The information identification model of the present invention combines 3D-CNN with Transformer, integrating the advantages of 3D-CNN in local feature extraction and Transformer in global information modeling, bringing unprecedented performance improvement to deep fake detection. The model is highly robust and efficient, and can still maintain stable and accurate detection effects when facing cross-compression and cross-library scenarios, that is, accurately identifying whether the face in the video has been deeply faked. Among them, cross-compression scenarios refer to the situation of evaluating model performance under different video compression rates. The deep fake detection model needs to maintain stable and accurate detection effects when facing videos with different compression rates. Therefore, the model needs to have good robustness and generalization capabilities.

[0041] In terms of model training, the present invention can adopt a multi-stage strategy. First, the UADFV dataset (Unrestricted Adversarial Deep Fake Video, a dataset containing real videos and fake videos) is used for preliminary training to learn basic deep fake features, and then the Face Forensics dataset (a large-scale video dataset specifically used to study image or video forgery detection) is used to optimize the model to enhance the model's expressiveness in complex scenes, for example, to enhance the model's ability to handle complex backgrounds, lighting changes, and videos of different resolutions. Second, the regularization techniques introduced, such as dropout (a method of randomly inactivating neurons), L2 regularization, etc., effectively prevent the model from overfitting and ensure the generalization ability of the model. For example, by introducing randomness and limiting the size of weights, the model effectively avoids overfitting of specific training samples, ensuring that the model can maintain robust performance on unseen data. Third, during the optimization and evaluation process, the present invention continuously monitors the training loss and validation set performance, fine-tunes hyperparameters such as learning rate and batch size, and ensures the stable performance of the model through cross-validation. Specifically, in the evaluation phase, multiple indicators (e.g., accuracy, precision, recall, and F1 score) are used to comprehensively measure the discriminative performance of the model.

[0042] Accuracy reflects the proportion of samples correctly classified by the model, precision measures the proportion of samples predicted by the model to be true that are actually true, recall calculates the proportion of samples that are actually true that are correctly predicted by the model, and the F1 score is the harmonic mean of precision and recall. These indicators can reflect the performance of the model.

[0043] The UADFV dataset contains a wealth of deep fake samples and is ideal for initial training and learning basic deep fake features. The Face Forensics dataset contains video data of different people, different environments, and different lighting conditions, which can help the model improve its performance in a wider range of scenarios. The present invention also performs standardization on the video in the preprocessing stage, such as face alignment, resizing, and format conversion, to ensure that each video can be effectively processed by the model.

[0044] The model of the present invention (i.e., the information identification model) can identify faces that have been deeply forged in videos with high accuracy, and can also demonstrate excellent generalization capabilities in different scenarios and resolutions.

[0045] In one embodiment, S102 may specifically include: using a target neural network model to process continuous frame data to obtain first difference information of the video to be identified, the first difference information being a difference parameter between each frame data within a preset time length; using a target structure model to process the first difference information to obtain long-term counterfeit information.

[0046] Although the target neural network model can simultaneously process the two-dimensional spatial information and time series information of the image, it cannot process long-term frame data. Therefore, the first difference information is a short-term difference feature, for example, the difference between X consecutive frame data. The target structure model can efficiently capture long-distance dependencies. The present invention will process the first difference information with the help of the target structure model to obtain long-term counterfeit information of the video to be identified.

[0047] Exemplarily, assuming that X is 3, the frame data are frame data 1, frame data 2, frame data 3, frame data 4, ..., frame data A, where A is an integer greater than or equal to 5. The first difference information may be difference information between three consecutive frame data, including but not limited to difference information between frame data 1 and frame data 3, difference information between frame data 1 and frame data 2, difference information between frame data 2 and frame data 3, difference information between frame data 1, frame data 2 and frame data 3, difference information between frame data 2 and frame data 4, difference information between frame data 3 and frame data 4, etc., and the present invention does not limit this. The long time sequence counterfeit information may be the difference between any two or more frame data between frame data 1 to frame data A.

[0048] On the one hand, the target neural network model is used to process continuous frame data to obtain the first difference information of the video to be identified, including: processing continuous frame data based on a preset duration to obtain at least one short-time frame data; using the target neural network model, determining the image difference information of at least one short-time frame data; based on the time mark of at least one short-time frame data, summarizing the image difference information of at least one short-time frame data to obtain the first difference information. The advantage of this setting is that the image difference information of each short-time frame data is arranged in sequence, so that the target structure model can parse the long-time forgery information.

[0049] Specifically, 3D-CNN is responsible for processing continuous frame data such as videos, and extracting the three-dimensional features and dynamic changes of faces from them. The 3D convolution kernel can simultaneously take into account the two-dimensional spatial information and time series information of the image, thereby capturing the spatiotemporal features of faces in the video. In the processing of each layer, the 3D convolution layer captures features at different depth levels through the deeply expanded convolution kernel, which can not only identify static facial features, but also capture dynamic changes, such as changes in subtle expressions such as blinking and smiling. These changes are difficult to fully imitate in deep fake videos and can be used to determine whether the video is fake information.

[0050] It is worth noting that the present invention will also determine the initial resolution of continuous frame data before processing the continuous frame data to obtain the first difference information of the video to be identified; adjust the initial resolution using image frame processing rules and facial image information of the continuous frame data to obtain the target resolution; and update the continuous frame data based on the target resolution.

[0051] Resolution is used to indicate the clarity of an image. The pre-set image frame processing rules can be understood as rules for blurring non-essential information and clarifying necessary information. Non-essential information refers to information that has little impact on authenticity verification, and necessary information refers to information that has a greater impact on authenticity verification. The purpose of this setting is to reduce the workload of the identification equipment while ensuring the accuracy of counterfeit information recognition, so that it does not need to process non-essential information, saving unnecessary resource consumption.

[0052] On the other hand, the first difference information is processed using the target structure model to obtain long-term counterfeit information, including: using the target structure model to process continuous frame data to obtain second difference information of the video to be identified, the second difference information is the difference parameter between all frame data; based on the second difference information and the first difference information, the long-term counterfeit information is determined.

[0053] Specifically, the second difference information can be understood as the long-term difference of the video to be identified obtained by the target structure model, and the long-term forgery information can be understood as the integration result of the second difference information and the first difference information. It can also be the use of the long-term difference information to adjust the short-term difference information, and the long-term difference information and the adjusted short-term difference information are summarized. The adjustment method includes: finding that one or some short-term differences are normal through the long-term difference. For example, the short-term difference information indicates that the eye sizes in frame data 1 and frame data 2 are different, but the long-term difference information indicates that the eye sizes are different because the eyes are closed. Comprehensively speaking, the eyes of the face in the video are normal, and then the difference is deleted from the short-term difference information.

[0054] Furthermore, the present invention can also directly process short-time difference information through the target structure model to obtain long-time forgery information. For example, 3D-CNN first processes the continuous frames in the video and extracts the feature map containing spatiotemporal information (i.e., the first difference information). Subsequently, these feature maps are input into the Transformer as input parameters, and the global change pattern of the dynamic face in the feature map is captured and understood through the self-attention mechanism of the Transformer, so as to obtain the long-time forgery information used to determine whether there is forged information in the video. In the present invention, 3D-CNN is responsible for refining the difference information existing in adjacent video frames, and passing the refined features to the Transformer, which further extracts the long-time forgery information. The advantage of this setting is that it can fully understand the dynamic changes of the face in the video, thereby accurately identifying deep forgery videos.

[0055] It is worth noting that in order to enable the model to run effectively on devices with limited resources, such as edge computing devices or mobile devices, the present invention will optimize and lightweight the model. Specifically, the present invention uses model compression techniques such as pruning, quantization, and knowledge distillation to reduce the number of model parameters and computational complexity without significantly sacrificing the performance of the model, so that the model can run more efficiently while maintaining high accuracy, thereby adapting to different hardware environments.

[0056] In order to integrate the model into actual application scenarios, the present invention also designs an application programming interface (API) interface, which enables the model to receive video data through network requests, perform deep fake detection, and return results. These API interfaces not only support common programming languages, but also can design user-friendly documents to reduce the difficulty of integration and simplify the integration process of the model.

[0057] S103: Process the long-term counterfeit information to obtain the authenticity identification result of the video to be identified.

[0058] Among them, the authenticity identification result of the video to be identified is used to indicate whether the face in the video to be identified is forged information. Specifically, if the face in the video to be identified is forged information, the forgery method and abnormal report can be further determined to assist enterprises or relevant institutions in locating the forger and blocking related forged information, thereby reducing the impact of forged information on enterprises and society.

[0059] Specifically, processing the long-term forged information can be understood as determining whether the long-term forged information has a pre-set abnormal evaluation item. If the abnormal evaluation item exists, it proves that the video to be identified is a forged video. The abnormal evaluation item can be the connection abnormality between two or more frame data, or the existence of a Class A abnormality in one or more frame data, which is not limited in the present invention.

[0060] In a specific implementation, the long-time sequence forged information can be a feature vector, and the long-time sequence forged information can be processed to obtain the authenticity identification result of the video to be identified, which can be executed by a classifier, that is, the long-time sequence forged information is input into the classifier for processing, and the authenticity identification result of the video to be identified is finally output. True means that the video to be identified is normal and there is no forged information. False means that the video to be identified is abnormal and the face in the video is forged. The forgery method includes but is not limited to replacement, partial feature change and synthesis. The present invention can also further determine the forgery method of the forged video in combination with abnormal factors, thereby providing data support for anti-counterfeiting work.

[0061] The technical solution of the above embodiment will analyze the video to be identified, obtain continuous frame data after face alignment, and then identify the difference features between the continuous frame data (i.e., long-term forged information), and determine whether the video to be identified has forged information (i.e., authenticity identification result) based on the long-term forged information. The forgery conclusion is determined based on the difference between each frame data, which is not limited to certain feature information. Both forged information with and without records can be identified. The identification method has high versatility and accuracy, and does not need to determine feature information such as blinking movements, pulse signals, eye reflection direction, and smoothness of facial lines. The detection efficiency is high, that is, it can accurately, comprehensively, and efficiently identify whether the face image in the video is forged information. The low versatility and detection efficiency of the feature detection method are solved. The model in the model detection method is trained with certain specific forged faces as training samples, which can well identify known forged information, but is difficult to identify unknown forged information.

[0062] Figure 2 is a flow chart of another method for identifying forged information provided by the present invention. Based on the above embodiments, this embodiment provides a preferred method for identifying forged information. Specifically, Figure 2 As shown, the method specifically includes:

[0063] S201. After receiving a video to be identified, use a preset face recognition model to process the video to be identified to obtain at least one frame of data and at least one frame of face image information.

[0064] S202: Processing facial image information of at least one frame of data using a preset face alignment model to obtain continuous frame data.

[0065] The continuous frame data is a video sequence of a single face image.

[0066] S203: Determine the initial resolution of continuous frame data.

[0067] S204: Using the image frame processing rules and the facial image information of the continuous frame data, the initial resolution is adjusted to obtain the target resolution.

[0068] S205: Update continuous frame data based on the target resolution.

[0069] S206: Process the continuous frame data using the target neural network model to obtain first difference information of the video to be identified.

[0070] The first difference information is a difference parameter between each frame data within a preset time length.

[0071] S207: Using the target structure model, the first difference information is processed to obtain long-term forged information.

[0072] S208: Process the long-term counterfeit information to obtain the authenticity identification result of the video to be identified.

[0073] The technical solution of the above embodiment will analyze the video to be identified, obtain continuous frame data after face alignment, and then identify the difference features between the continuous frame data (i.e., long-term forged information), and determine whether the video to be identified has forged information (i.e., authenticity identification result) based on the long-term forged information. The forgery conclusion is determined based on the difference between each frame data, which is not limited to certain feature information. Both forged information with and without records can be identified. The identification method has high versatility and accuracy, and does not need to determine feature information such as blinking movements, pulse signals, eye reflection direction, and smoothness of facial lines. The detection efficiency is high, that is, it can accurately, comprehensively, and efficiently identify whether the face image in the video is forged information. The low versatility and detection efficiency of the feature detection method are solved. The model in the model detection method is trained with certain specific forged faces as training samples, which can well identify known forged information, but is difficult to identify unknown forged information.

[0074] Figure 3 Schematic diagram of the structure of a device for identifying counterfeit information provided by the present invention. Figure 3 As shown, the device includes: an acquisition module 301, a determination module 302 and an identification module 303.

[0075] The acquisition module 301 is used to determine target feature data of the video to be identified, wherein the target feature data is continuous frame data, and the frame data includes face image information.

[0076] The determination module 302 is used to process the continuous frame data using the pre-trained information identification model to obtain the long-term forgery information of the video to be identified, wherein the long-term forgery information is used to indicate the difference between the continuous frame data.

[0077] The identification module 303 is used to process the long-term forged information to obtain the authenticity identification result of the video to be identified.

[0078] Optionally, the acquisition module 301 is specifically used to, after receiving the video to be identified, process the video to be identified using a preset face recognition model to obtain at least one frame data and at least one frame data of face image information; use a preset face alignment model to process the face image information of at least one frame data to obtain continuous frame data, wherein the continuous frame data is a video sequence of a single face image.

[0079] Optionally, the information identification model includes a target neural network model and a target structure model.

[0080] Optionally, the determination module 302 is specifically used to use the target neural network model to process continuous frame data to obtain first difference information of the video to be identified, wherein the first difference information is the difference parameter between each frame data within a preset time length; and use the target structure model to process the first difference information to obtain long-term counterfeit information.

[0081] Optionally, the determination module 302 is also used to determine the initial resolution of the continuous frame data before processing the continuous frame data to obtain the first difference information of the video to be identified; adjust the initial resolution using the image frame processing rules and the facial image information of the continuous frame data to obtain the target resolution; and update the continuous frame data based on the target resolution.

[0082] Optionally, the determination module 302 is specifically used to process continuous frame data based on a preset time length to obtain at least one short-time frame data; use the target neural network model to determine the image difference information of at least one short-time frame data; based on the time identifier of at least one short-time frame data, summarize the image difference information of at least one short-time frame data to obtain first difference information.

[0083] Optionally, the determination module 302 is specifically used to use the target structure model to process continuous frame data to obtain second difference information of the video to be identified, wherein the second difference information is the difference parameter between all frame data; based on the second difference information and the first difference information, determine the long-term counterfeit information.

[0084] The device for identifying forged information provided in the above embodiments can execute the method for identifying forged information provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0085] Figure 4 : is a schematic diagram of the structure of an electronic device provided by the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0086] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (also called random access memory, RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, ROM 12 and RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0087] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0088] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs the various methods and processes described above, such as a method for identifying counterfeit information.

[0089] In some embodiments, the method for identifying counterfeit information may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for identifying counterfeit information described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the method for identifying counterfeit information in any other appropriate manner (e.g., by means of firmware).

[0090] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0091] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0092] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0093] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0094] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server) or a computing system that includes middleware components (e.g., an application server) or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein) or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0095] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0096] In one embodiment, the present invention further includes a computer program product. The computer program product includes a computer program. When the computer program is executed by a processor, the method for identifying counterfeit information according to any embodiment of the present invention is implemented.

[0097] The computer program product may be implemented in a computer program code for performing the operation of the present invention written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​and conventional procedural programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0098] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0099] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying forged information, characterized in that: include: Determine target feature data of the video to be identified, wherein the target feature data is continuous frame data, and the frame data includes face image information; Using a pre-trained information identification model, the continuous frame data is processed to obtain long-term forgery information of the video to be identified, wherein the long-term forgery information is used to indicate the difference between the continuous frame data; The long-term forged information is processed to obtain an authenticity identification result of the video to be identified.

2. The method according to claim 1, characterized in that The step of determining target feature data of the video to be identified includes: After receiving the video to be identified, the video to be identified is processed using a preset face recognition model to obtain at least one frame of data and face image information of the at least one frame of data; The face image information of the at least one frame data is processed using a preset face alignment model to obtain the continuous frame data, wherein the continuous frame data is a video sequence of a single face image.

3. The method according to claim 1, characterized in that The information identification model includes a target neural network model and a target structure model. The pre-trained information identification model is used to process the continuous frame data to obtain the long-time sequence forgery information of the video to be identified, including: Using the target neural network model, the continuous frame data is processed to obtain first difference information of the video to be identified, wherein the first difference information is a difference parameter between each frame data within a preset time length; The first difference information is processed using the target structure model to obtain the long-term forged information.

4. The method according to claim 3, characterized in that: Before the continuous frame data is processed to obtain the first difference information of the video to be identified, the method further includes: Determining an initial resolution of the continuous frame data; Using the image frame processing rules and the facial image information of the continuous frame data, the initial resolution is adjusted to obtain a target resolution; Based on the target resolution, the continuous frame data is updated.

5. The method according to claim 3, characterized in that: The method of using the target neural network model to process the continuous frame data to obtain the first difference information of the video to be identified includes: Processing the continuous frame data based on the preset duration to obtain at least one short time sequence frame data; Determining image difference information of the at least one short time series frame data using the target neural network model; Based on the time mark of the at least one short-time sequence frame data, the image difference information of the at least one short-time sequence frame data is summarized to obtain the first difference information.

6. The method according to claim 4, characterized in that The step of processing the first difference information by using the target structure model to obtain the long time series forgery information includes: Using the target structure model, the continuous frame data is processed to obtain second difference information of the video to be identified, wherein the second difference information is a difference parameter between all frame data; The long time series falsification information is determined based on the second difference information and the first difference information.

7. A device for identifying forged information, characterized in that: Used to implement the method for identifying forged information as described in any one of claims 1 to 6, the device for identifying forged information comprises: An acquisition module, used to determine target feature data of a video to be identified, wherein the target feature data is continuous frame data, and the frame data includes face image information; A determination module, used to process the continuous frame data using a pre-trained information identification model to obtain long-term forgery information of the video to be identified, wherein the long-term forgery information is used to indicate the difference between the continuous frame data; The identification module is used to process the long-time sequence forgery information to obtain the authenticity identification result of the video to be identified.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for identifying counterfeit information as claimed in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for identifying counterfeit information as claimed in any one of claims 1 to 6 when executed.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the method for identifying counterfeit information according to any one of claims 1 to 6.

Images