Authenticity assessment of the modified content

Abstract

A system for evaluating the authenticity of modified content includes a computing platform having a hardware processor and a memory storing software code that includes a neural network. The neural network is trained to evaluate the authenticity of modified content generated based on baseline digital content, including one or more modifications made to the baseline digital content. The hardware processor executes the software code to use the neural network to receive the modified content and evaluate the authenticity of each of the one or more modifications made to the baseline digital content, generating one or more authenticity evaluations corresponding to the one or more modifications made to the baseline digital content, respectively. The hardware processor is further configured to execute the software code to generate an authenticity evaluation of the modified content based on the one or more authenticity evaluations and output the authenticity evaluation for presentation on a display.

Description

[0001] Cross - Reference to Related Applications

[0002] This application is related to U.S. Patent Application No. 16 / 737,810, entitled "Content Verification Based on Intrinsic Attributes" (Attorney Docket No. 0260635), which was filed concurrently with this application and the entire content of which is incorporated herein by reference. Background of the Invention

[0003] Advances in machine learning have enabled the creation of realistic but fake reproductions of a person's image or voice, known as "deepfakes," through the use of deep artificial neural networks. Deepfakes can be created without the consent of the person whose image or voice is used and may make the represented person appear to have said or done things they actually did not say or do. Thus, digitally manipulated content using deepfakes can be maliciously used to spread misinformation.

[0004] Due to the widespread popularity of digital content in entertainment and news distribution, effective verification and management of such content are important for its creators, owners, and distributors. However, with the continuous improvement of machine - learning solutions, deepfakes are and will continue to be difficult to detect. Thus, in violation of contractual agreements or regulatory restrictions, digitally manipulated or even completely fake content may be inadvertently played or otherwise distributed, exposing content owners and / or distributors to potential legal risks. Summary of the Invention

[0005] Systems and methods for assessing the authenticity of modified content are provided herein, generally shown and / or described in at least one of the related drawings and more fully set forth in the claims. Brief Description of the Drawings

[0006] Figure 1 An exemplary schematic diagram is shown according to one embodiment, in which baseline digital content has been modified by one or more users;

[0007] Figure 2 A schematic diagram of an exemplary system for assessing the authenticity of modified content is shown according to one embodiment;

[0008] Figure 3 An exemplary schematic diagram of software code is shown according to one embodiment, which includes a neural network (NN) trained to assess the authenticity of modified content;

[0009] Figure 4 A flowchart is shown according to one embodiment, which presents an exemplary method for assessing the authenticity of modified content;

[0010] Figure 5Shows an exemplary graphical user interface (GUI) of a system for evaluating the authenticity of modified content according to one embodiment. Detailed Description

[0011] The following description contains specific information about the implementations in this disclosure. Those skilled in the art will recognize that the present disclosure can be implemented in ways different from those specifically discussed herein. The accompanying drawings and the accompanying detailed description of this application are directed only to exemplary embodiments. Unless otherwise stated, the same or corresponding elements in the drawings are identified by the same or corresponding reference numerals. Moreover, the illustrations and appendices in this application Figure 1 generally are not to scale and are not intended to correspond to actual relative sizes.

[0012] This application discloses systems and methods for evaluating the authenticity of modified content, which overcome the disadvantages and deficiencies in the prior art. At the original or "baseline" version of the digital content, by using a trained neural network to evaluate the authenticity of one or more digital content modifications, this application discloses an ingenious verification solution that allows for a detailed evaluation of the modified content, which improves the prior art. In addition, by generating verification data for evaluating the modified content based on the inherent attributes of one or more modifications made to the underlying baseline digital content, this solution conveniently utilizes the characteristics of each modification to evaluate its authenticity.

[0013] It is noted that in some embodiments, the content verification solution can be executed by a basic automation system as a basic automation process. It is noted that the terms "automation", "automated", and "automate" used in this application refer to systems and processes that do not require the participation of a human user (such as a system administrator). However, in some embodiments, a human system operator or administrator may receive the verification confirmation information generated by the automated system, and according to the automated methods described herein, human participation is optional. Therefore, the methods described in this application can be executed under the control of the hardware processing components of the disclosed automated system.

[0014] Figure 1 Shows illustration 100 in a demonstration use case according to one embodiment, in which baseline digital content 122 is modified by user 102b during an authorized modification process, and can be further modified by user 102c in an authorized or unauthorized modification process. Figure 1 Includes user 102a using device 110a, user 102b using device 110b, user 102c using device 110c, and user 102d using device 110d. Figure 1Also shown are examples of modified content 120a and further modified content 102b. It is noted that the "modified content" as defined in this application refers to the baseline digital content 122, which has been modified one or more times after being created. In addition, the "authentic content" as defined herein can refer to either the baseline digital content that has not been modified or the modified content that only includes the modifications introduced by the authorized users of the content (such as the creator of the content or the user authorized to make modifications to the baseline digital content).

[0015] It is noted that the baseline digital content 122 can be determined in any suitable manner like this, including but not limited to the methods described in U.S. Patent Application 16 / 737,810 titled "Content Verification Based on Intrinsic Attributes", which is incorporated herein by reference in its entirety as described above. For the purposes of this application, the baseline digital content and / or further modified content 120b included in the modified content 120a precisely refers to the content marked or described therein.

[0016] According to Figure 1 In the illustrated embodiment, user 102b can receive the baseline digital content 122 from user 102a. For example, user 102b can receive the baseline digital content 122 from a device 110a controlled by user 102a. User 102a is the creator of the baseline digital content 122 or an authorized provider of the baseline digital content 122. User 102a can use, for example, one or more microphones, and / or one or more depth sensors, and / or one or more cameras, and / or one or more still image cameras integrated with the device 110a to generate the baseline digital content 122. In various embodiments, the baseline digital content 122 can be video content without audio, audio content without video, or video-audio content of a movie, episodic television (TV) content, to name a few, which may include web series and / or video logs, sports content, news content, or video game content. Alternatively, in some embodiments, the baseline digital content 122 can be in the form of a digital photograph.

[0017] Note that the narrative of Device 100a as a camera, Device 110b as a smartphone, Device 110c as a laptop, and Device 110d as a desktop computer is merely exemplary. In many other embodiments, one or more of Devices 110a, 110b, 110c, and 110d (hereinafter referred to as "Devices 110a-110d") may take the form of a tablet computer, a game console, a smart TV (Smart TV), or a wearable communication device. Moreover, when implemented as a wearable communication device, Devices 100a-100d can take the form of augmented reality (AR) or virtual reality (VR) headsets or glasses, smartwatches, smart rings, or any other smart personal item carried by Users 102a, 102b, 102c, or 102d (hereinafter referred to as "Users 102a-102d"), or located inside or on the clothing worn by Users 102a-102d.

[0018] As Figure 1 shown, User 102b uses Device 110b to receive the baseline digital content 122 of Device 110a controlled by User 102a, and can use Device 110b to modify the baseline digital content 122 to generate modified content 120a. According to Figure 1 the exemplary implementation shown, the modification of the baseline digital content 122 is an authorized modification. In some embodiments, as Figure 1 shown, the modifications made by Device 102b and Device 110b to the baseline digital content 122 cause the edit data 121a describing the authorized modification to be transmitted to and stored in the secure transaction ledger 101, and used in the subsequent verification of the modified content 120a. Note that in many embodiments, the secure transaction ledger 101 can take the form of a public or private secure transaction ledger. By way of example, instances of the secure transaction ledger can include blockchains, hashgraphs, directed acyclic graphs (DAGs), and Holochain ledgers, among others.

[0019] As Figure 1 further shown, User 102c can use Device 110c to receive the modified content 120a from Device 110b, and can use Device 110c to further modify the modified content 120a to generate further modified content 120b. In some embodiments, the further modification of the modified content 120a by User 102c is authorized, and the edit data 121b describing the authorized further modification is also transmitted to and stored in the secure transaction ledger 101, and used in the subsequent verification of the modified content 120b. However, when the further modification of the modified content 120a is not authorized, the edit data 121b is not generated and not stored in the secure transaction ledger 101.

[0020] AlthoughFigure 1 The illustrated embodiment describes that user 102b makes an authorized modification to the baseline digital content 122 and then forwards the modified data 120a to user 102c. This use case is merely exemplary. In other examples, user 102c can directly receive the baseline digital content 122 from user 102a, as shown by the dashed line 111 directly connecting devices 110a and 110c. That is, in some embodiments, the baseline digital content 122 may not have undergone an authorized modification by user 102b before an unauthorized modification by user 102c.

[0021] The modification of the baseline digital content 122 and the modified content 120a can be performed by using one or more content processing applications, such as audio and / or video recording or editing applications stored in devices 110b and 110c. Further modification of the modification of the baseline digital content 122 or the modified content 120a can include modification of one or more segments of the audio content and / or video content in the baseline digital content 122 and / or the modified content 120a. The modification of the baseline digital content 122 or the modified content 120a can include only modification of the video content, modification of the audio content, modification of immersive content (such as deep point clouds for virtual reality applications), or include modification of any combination of audio, video, and immersive content.

[0022] When the modification modifies the video content of the baseline digital content 122 or the modified content 120a, the modification can include changing the image attributes (e.g., contrast, brightness, etc.) of one or more video frames, deleting one or more video frames, inserting one or more video frames, removing objects from one or more video frames, inserting objects into one or more video frames, changing the color of one or more video frames, or adding one or more of the metadata to one or more video frames. Similarly, when the modification modifies the baseline digital content 122 in the form of a digital picture, the modification can include changing the image attributes (e.g., contrast, brightness, etc.) of the digital picture, removing objects from the digital picture, inserting objects into the digital picture, changing the color of the digital picture, or adding one or more of the metadata to the digital picture.

[0023] When the modification of the baseline digital content 122 or the modified content 120a modifies the audio content of the baseline digital content 122 or the modified content 120a, the modification may include deleting a part of the original audio content and / or inserting additional audio content, such as music or speech. Optionally, the operations performed on the audio content of the baseline digital content 122 or the modified content 120a may include mixing audio tracks, changing the audio volume of one or more audio tracks, or adding metadata to one or more audio tracks. When the modification of the baseline digital content 122 or the modified content 120a modifies immersive content, such as a deep point cloud captured by the device 110a using a virtual reality application, for example, the modification may include deleting a part of the deep point cloud.

[0024] According to Figure 1 the illustrated embodiment, the user 102d uses the device 110d to receive the modified content 120a (generated by the user 102b) from the device 110b, or the further modified content 120b (if generated by the user 102c) from the device 110c. However, lacking a sound solution for evaluating the authenticity of the modification to the baseline digital content 122 or the further modification to the baseline digital content 120a, the user 102d cannot guarantee that the modified content 120a or the further modified content 120b only contains authorized modifications to the baseline digital content 122. Therefore, the uncertainty resulting from the lack of a sound solution for evaluating the authenticity of the modified content may hinder the supply, consumption, and enjoyment of authentic content, which is not desirable.

[0025] Figure 2 FIG. shows a schematic diagram of an exemplary system 230 for evaluating the authenticity of modified content according to one embodiment. As described below, the system 230 can be implemented by using a computer server accessible on a local area network (LAN), or can be implemented as a cloud-based system. Moreover, as Figure 2 shown, the system 230 includes a computing platform 232 having a hardware processor 234, and a system memory 236 implemented as a non-transitory storage device. According to this exemplary embodiment, the system memory 236 stores software code 240 that provides a graphical user interface (GUI) 238, and the graphical user interface 238 includes an authenticity assessment 270 configured to evaluate the authenticity of the modified content 220. As Figure 2 further shown, the modified content 220 includes a baseline digital content 222, verification data 224 for the baseline digital content 222, and a modification 226 to the baseline digital content 222. According to Figure 2 the illustrated exemplary embodiment, the modification 226 to the baseline digital content 222 includes a modification 226a having verification data 228, and a modification 226b lacking the verification data.

[0026] According to some embodiments, each of modification 226a and modification 226b can represent all the modifications made to the baseline content 222 in the modification content 220. That is, modification 226a can correspond to all the modifications with verification data 228, while modification 226b can correspond to all the modifications lacking verification data. However, in other embodiments, each individual modification to the baseline content 222, such as the application of a sepia filter or other color change, a change in brightness, etc., can be stored as a separate modification.

[0027] For example, each modification to the baseline content 222 can be stored separately and can be bound together during distribution or playback. For example, in one embodiment, an editing application can be configured to present the modified content when generating a "reverse graphics processing software document (.psd) file", where this "reverse.psd file" has instructions to perform a transformation that negates all the changes applied to the baseline content 222 during presentation, so that the baseline content 222 included in the modification content 220 can be saved without loss. In one embodiment, an extension of the Exchangeable Image File Format (EXIF) tag can be used to record the modification. The editing application can be configured to embed the "reverse.psd file" such that another user receiving the modification content 220 can choose to extract it and execute it to regenerate the baseline content 222. In another embodiment, the editing application can be configured to determine whether the baseline content 222 has been imported. If it has been imported, the application can export the baseline content 222 and modifications 226a and 226b separately.

[0028] Similarly, as Figure 2 shown, the system 230 is implemented in a usage environment that includes a secure transaction ledger 201, a device 210d including a display 212, and a user 202d using the device 210d. It is noted that the communication network 204 and the network communication link 206 communicatively connect the system 230 to the device 210d, and optionally, the system 230 can be connected to the secure transaction ledger 201 so that the system 230 can receive the modification content 220 from the device 210d and output an authenticity assessment 270 for presentation on the display 212 of the device 210d.

[0029] The modification content 220 containing the baseline digital content 222 generally corresponds to either or both of the modification content 120a and the further modification content 120b, both of which contain the baseline digital content 122. As a result, the underlying digital content 222 and the modification content 220 can share any corresponding features attributed to the underlying digital content 122 and either or both of the modification content 120a and the further modification content 120b, and vice versa.

[0030] In addition, user 202d, device 210d including display 212, and security ledger 201 generally correspond to user 102d, device 110d, and security ledger 101 in Figure 1 respectively. As a result, user 202d, device 210d, and security ledger 201 can share any features of the present disclosure attributable to the corresponding ones of user 102d, device 110d, and security ledger 101, and vice versa. That is, although device 210d is shown as a desktop computer in Figure 2 , this illustration is provided only as an example. More generally, device 210d can be any suitable mobile or stationary computing device or system that implements data processing capabilities sufficient to provide a user interface, support a connection to communication network 204, and implement the functions of device 210d.

[0031] It should be noted that, in various embodiments, display 212 can be physically integrated with device 210d, or can be communicatively connected to device 210d but physically separated from device 210d. For example, in the case where device 210d is implemented as a smartphone, laptop, or tablet, display 212 will typically be integrated with device 210d. In contrast, in the case where device 210d is implemented as a desktop computer, display 212 can take the form of a monitor separated from device 210d in the form of a computer tower. More notably, display 212 can be implemented as a liquid crystal display (LCD), light-emitting diode (LED) display, organic light-emitting diode (OLED) display, or any other suitable display screen that performs the physical conversion of signals to light.

[0032] Regarding system 230, although this application refers to software code 240 stored in system memory 236 for the sake of conceptual clarity, more generally, system memory 236 can take the form of any computer-readable non-transitory storage medium. The expression "computer-readable non-transitory storage medium" used in this application refers to any medium other than a carrier wave or other transitory signal that provides instructions to the hardware processor 234 of computing platform 232. Thus, computer-readable non-transitory media can correspond to various types of media, such as volatile media and non-volatile media. Volatile media can include dynamic memory, such as dynamic random access memory (dynamic RAM), while non-volatile memory can include optical, magnetic, or electrostatic storage devices. Common forms of computer-readable non-transitory media include, for example, optical discs, RAM, programmable read-only memory (PROM), erasable PROM (EPROM), and flash memory.

[0033] It should also be noted that although Figure 2The software code 240 is described as being stored in its entirety in the system memory 236, but this representation is provided merely for conceptual clarity. More generally, the system 230 may include one or more computing platforms 232, such as multiple computer servers, which may be co-located or may form an interconnected but distributed system, such as a cloud-based system.

[0034] As a result, the hardware processor 234 and the system memory 236 correspond to distributed processor and memory resources within the system 230. Accordingly, it should be understood that various features of the software code 240, such as one or more of the features described below with reference to Figure 3 can be stored and / or executed using the distributed memory and / or processor resources of the system 230.

[0035] According to Figure 2 the illustrated embodiment, the user 202d can interact with the system 230 via the communication network 204 using the device 210d to evaluate the authenticity of the modification 220. In the illustrated embodiment, the computing platform 232 may correspond to one or more web servers and may be accessible, for example, via a packet-switched network such as the Internet. Alternatively, the computing platform 232 may correspond to one or more computer servers supporting a wide area network (WAN), a local area network (LAN), or included in another type of limited distribution or private network.

[0036] Figure 3 FIG. shows an exemplary schematic diagram of software code 340 according to one embodiment, the software code 340 including a neural network 342 that is trained to evaluate the authenticity of modifications 120a / 120b / 220 generated based on Figure 1 and Figure 2 the baseline digital content 122 / 222 in and one or more modifications 226 made to the baseline digital content 122 / 222. It should be noted that, as defined in the present application, an artificial neural network, also simply referred to as a "neural network" (hereinafter referred to as "NN"), is a machine learning framework in which patterns or learned representations of observational data are processed using highly connected computational layers that map the relationship between inputs and outputs.

[0037] In the context of deep learning, a "deep neural network" can refer to a neural network that utilizes multiple hidden layers between an input layer and an output layer, which can allow learning based on features not explicitly defined in the original data. Thus, various forms of neural networks, including but not limited to deep neural networks, can be used to make predictions on new data based on past examples or "training data". In various embodiments, a neural network can be utilized to perform image analysis and / or natural language analysis. It is noted that the training data can be archived data, or content previously created or modified by one of users 102a or 102b, for example, and the authenticity of the content creation or modification has been verified. In those use cases, the neural network can learn the user's creation or editing tendencies, such as the devices and / or applications that each user typically uses to create or modify content, as identified by metadata embedded or otherwise included in the data file containing the content. A substantial deviation from those learned tendencies can be used to flag the content or a modification to the content as suspicious.

[0038] As Figure 3 shown, in addition to neural network 342, software code 340 includes an evaluation module 346 and can include an optional scoring module 344. As Figure 3 further shown, neural network 342 of software code 340 is configured to receive modified content 320 as input and provide one or more authenticity assessments 352a and / or 352b as output. In some embodiments, evaluation module 346 can be configured to generate authenticity assessment 370 based on authenticity assessment 352a and / or authenticity assessment 352b.

[0039] However, in embodiments where software code 340 includes scoring module 344, neural network 342 can also output authenticity assessment 352a and / or authenticity assessment 352b to scoring module 344, and scoring module 344 can in turn provide one or more of evaluation confidence score 354a corresponding to authenticity assessment 352a and evaluation confidence score 354b corresponding to authenticity assessment 352b to evaluation module 346. In those embodiments, evaluation module 346 can be configured to generate authenticity assessment 370 based on evaluation confidence score 354a and / or evaluation confidence score 354b and authenticity assessment 352a and / or authenticity assessment 352b.

[0040] Modified content 320 generally corresponds Figure 2 to modified content 220 in Figure 1 and one or both of modified content 120a and further modified content 120b in Figure 2The authenticity assessment 270 therein. Thus, the authenticity assessment 370 can share any features of the present disclosure that belong to the authenticity assessment 270, and vice versa. Moreover, the modification content 320 can share any features of the present disclosure that belong to the modification content 220, the modification content 120a, or the further modification content 120b, and vice versa. That is to say, like the modification content 220, the modification content 320 can include the baseline digital content 222, the verification data 224 for the baseline digital content 222, the modification 226a made to the baseline digital content 222 with the verification data 228, and the modification 226b made to the baseline digital content 222 without the said verification data.

[0041] In addition, the software code 340 generally corresponds to the software code 240, and the corresponding features in the present disclosure can be shared arbitrarily. Therefore, like the software code 340, the software code 240 can include a neural network corresponding to the neural network 342, and multiple features respectively corresponding to the evaluation module 346 and the optional scoring module 344 in some embodiments.

[0042] will be combined with Figure 2 、 Figure 3 and Figure 5 and with reference to Figure 4 further describe the functions of the software code 240 / 340. Figure 4 FIG. 460 shows a flowchart presenting an exemplary method for evaluating the authenticity of modification content according to one embodiment. Regarding Figure 4 the method outlined in

[0043] Figure 5 it is worth noting that certain details and features have been omitted from the flowchart 460 so as not to obscure the discussion of the inventive features in this application.

[0043] Figure 5 FIG. 538 shows an exemplary graphical user interface provided by the software code 240 / 340 of the system 230 according to one embodiment. As Figure 5 shown, the exemplary graphical user interface 538 provides an authenticity assessment 570 to identify the baseline digital content 522, and includes a modification pane 572 and an optional visualization 574 in the form of a demonstration heat map representing the modification content 520. However, in other embodiments, the optional visualization 574 can take forms such as a table, a list, a badge, or an icon to distinguish the unmodified part of the baseline digital content and the modification of the baseline digital content. As Figure 5 further shown, the modification pane 572 identifies the modifications 526a and 526b of the baseline digital content 522, and the authenticity assessments 552a and 552b respectively corresponding to at least the modifications 526a and 526b. In Figure 5In an exemplary embodiment, the shown modification pane 572 also includes evaluation confidence scores 554a and 554b based on respective authenticity evaluations 552a and 552b.

[0044] The authenticity evaluation 570 generally corresponds Figure 2 and Figure 3 to the authenticity evaluations 270 / 370, and the graphical user interface 538 generally corresponds Figure 2 to the graphical user interface 238 in. That is, the graphical user interface 238 and the authenticity evaluations 270 / 370 can share any features attributed to the corresponding graphical user interface 538 and authenticity evaluation 570 in this disclosure, and vice versa. Thus, although Figure 2 and Figure 3 are not shown, the authenticity evaluations 270 / 370 can identify the baseline digital content 222 / 522 and can have multiple features corresponding to the modification pane 572 and the optional visualization 574 respectively.

[0045] The baseline digital content 522 generally corresponds Figure 1 and Figure 2 to the baseline digital content 122 / 222. Thus, the baseline digital content 522 can share any features attributed to the baseline digital content 122 / 222 in this disclosure, and vice versa. Moreover, the modification content 520 generally corresponds Figure 2 and Figure 3 to the modification content 220 / 320, and Figure 1 either or both of the modification content 120a and the further modification content 120b. Thus, the modification content 520 can share any features attributed to the modification content 220 / 320, the modification content 120a, or the further modification content 120b in this disclosure, and vice versa. That is, like the modification content 220, the modification content 520 can include the baseline digital content 222 / 522, the verification data 224 for the baseline digital content 222 / 522, the modifications 226a / 526a made to the baseline digital content 222 / 522 with the verification data 228, and the modifications 226b / 526b made to the baseline digital content 222 / 522 without the verification data.

[0046] It should be emphasized that Figure 2 , Figure 3 and Figure 5The modified content 220 / 320 / 520 is represented as including two modifications to the baseline digital content 222 / 522, namely modifications 226a / 526a and 226b / 526b, solely for conceptual clarity. In some embodiments, the modified content 220 / 320 / 520 may include a single modification to the baseline digital content 222 / 522, while in other embodiments, the modified content 220 / 320 / 520 may include more than two modifications to the baseline digital content 222 / 522.

[0047] According to this exemplary embodiment, the authenticity of each of the modifications 226a / 526a and 226b / 526b is the subject of an assessment, resulting in authenticity assessments 270 / 370 / 570. In some embodiments, as Figure 5 shown, the authenticity assessments 270 / 370 / 570 include visualizations 574 that graphically depict the authenticity assessments 352a / 552a and 352b / 552b of the corresponding modifications 226a / 526a and 226b / 526b to the baseline digital content 222 / 522. For example, as Figure 5 further shown, in some embodiments, the visualization 574 may take the form of a heat map of the modified content 220 / 320 / 520 based on the verification assessments 352a / 552a and / or the verification assessments 352b / 552b.

[0048] In some embodiments, the authenticity assessments 270 / 370 / 570 may further include assessment confidence scores 354a / 554a determined based on the authenticity assessments 352a / 552a and / or assessment confidence scores 354b / 554b determined based on the authenticity assessments 352b / 552b. In the above embodiments, the visualization 574 may take the form of a heat map of the modified content 220 / 320 / 520 based on the assessment confidence scores 354a / 554a and / or the assessment confidence scores 354b / 554b.

[0049] Referring to the exemplary heat map visualization 574, the modified content 520 is divided into segments 580a, 580b, 580c, 580d, 580e, and 580f (hereinafter referred to as "segments 580a - 580f"), arranged in sequence along the horizontal x-axis, which is consistent with the time code 576 of the modified content 520. In addition to segments 580a - 580f, the heat map visualization 574 includes the modifications 526a and 526b of the corresponding authenticity assessments 552a and 552b. That is, in Figure 3In it, the segment 580a of the baseline digital content 522 is modified by the modification 526a in the modified content 520, and a visual description of the authenticity assessment 552a of the modification 526a determined by the neural network 342 of the software code 340 is shown. Additionally, the segments 580d and 580e of the baseline digital content 522 are modified by the modification 526b in the modified content 520, and a visual description of the authenticity assessment 552b of the modification 526b determined by the neural network 342 of the software code 340 is shown.

[0050] It should be noted that although the heatmap visualization 574 includes the timecode 576, this representation is merely exemplary, and in particular, it can relate to use cases where the baseline content 522 or the modified content 520 takes the form of a video segment or an audio track. In embodiments where the baseline content 522 or the modified content 520 takes the form of a digital photograph omitting the timecode 576, the heatmap visualization 574 can be used to highlight regions based on the likelihood that the regions within the digital photograph have been modified. The determination of the likelihood that a particular region of the digital photograph has been modified can be based on changes compared to other images created by the same user, device, or application used to create the baseline content 522, or based on changes compared to other images created at the same location and / or time as the baseline content 522.

[0051] It should also be noted that the segmentation of the modified content 520 into segments 580a - 580f can occur as part of the authenticity assessment of the modified content 520, or occur naturally due to the nature of the baseline content 522 itself. For example, when the baseline content 522 includes a video segment that includes the timecode 576 generated during the creation of the baseline content 522, the same original timecode 576 can be used to force-segment the modified content 520 so that the content included in the timecode increments or segments can be compared between the baseline content 522 and the modified content 520. Or, when assessing the authenticity of the modified content 520, a completely new segmentation scheme can be forced on the baseline content 522 and the modified content 520, such as according to timecode intervals or frame sequences. Additionally, in some embodiments, more than one segmentation scheme can be applied during the assessment of the modified content 520.

[0052] According to Figure 5In the exemplary embodiment shown, heat map visualization 574 simultaneously shows all segments of the baseline digital content that have been modified in the modification 520 to facilitate comparison of their respective authenticity assessments. However, in other embodiments, it may be advantageous or desirable to display all or fewer than all of the segments included in the modification 520. For example, in various embodiments, the visualization 574 may display all segments of the modification 520, only the segments that have been modified, only the segments that include or do not include verified modifications, segments selected by the user, and so on.

[0053] It should be noted that each segment 580a - 580f can be visually distinguished by its fill pattern or darkness in Figure 5 . In some embodiments, color, hue, or the absence of color (instead of fill pattern or darkness) can also be used to make visual distinctions between the segments 580a - 580f. That is, for example, the substantially colorless or "white" segments 580b, 580c, and 580f of the modification 520 may remain unchanged relative to their corresponding segments in the baseline digital content 522. Again, by way of example only, the segment 580a of the modification 520 that includes the verified modification 526a to the baseline digital content 522 can be described with a color such as green, indicating that the broadcast or distribution is secure. In contrast, still by way of example only, the segments 580d and 580e of the modification 520 that include the unverified modification 526b of the baseline digital content 522 can be described with a color such as red to give a warning signal.

[0054] It should also be noted that although in some embodiments, the modified segments 580a, 580d, and 580e can be colored to correspond to their respective authenticity assessments, in other embodiments, it may be advantageous or desirable to color the modified segments 580a, 580d, and 580e based on the degree to which they are changed compared to the corresponding segments of the baseline digital content 522. In such embodiments, for example, the unmodified segments of the modification 520 can be colored green, the mildly modified segments can be colored yellow, and the severely modified segments or segments that have undergone multiple layers of modification can be colored red.

[0055] In some embodiments, different weights can be assigned to different types of modifications. For example, aspect ratio changes and compression are less likely to be the result of malicious or unauthorized modifications compared to brightness changes or green screen visual effects. Thus, a mildly modified segment can be identified based on its including a single low - risk type of modification, while a severely modified segment can be identified based on multiple modifications and / or a single high - risk modification.

[0056] For example, each of the segments 580a - 580f can be a time code interval of the modified content 520. That is, in some embodiments, the modified content 520 can be segmented based on the time code 576 of the modified content 520. Alternatively, the segments 580a - 580f can be any other predetermined content segments suitable for the nature of the baseline digital content 522 and the modified content 520. For example, in the case where the baseline digital content 522 and the modified content 520 include audio, the segments 580a - 580f can each take the form of an audio track of the modified content 520. In the case where the baseline digital content 522 and the modified content 520 include video, the segments 580a - 580f of the modified content 520 can correspond to one or more video frames, one or more video "shots", or one or more video scenes.

[0057] It should be noted that, as used in this application, a "shot" refers to a sequence of video frames captured from a unique camera perspective without cuts and / or other cinematic transitions. Thus, in various embodiments, each of the segments 580a - 580f of the modified content 520 can correspond to a single video frame, a single video shot containing multiple individual video frames, or one or more scenes including multiple shots.

[0058] In embodiments where the software code 240 / 340 includes a scoring module 344, the heat map visualization 574 can also display an evaluation confidence score 554a based on the authenticity assessment 552a and an evaluation confidence score 554b based on the authenticity assessment 552b. The evaluation confidence scores 554a and 554b can be represented numerically, for example, numbers in the range from 0 (no confidence) to 100 (maximum confidence). In Figure 5 an example, the evaluation confidence score 554a is represented numerically as 90, while the evaluation confidence score 554b is represented numerically as 80. In another example, the evaluation confidence scores 554a and 554b can be represented graphically, for example, by bars with different widths or heights. In Figure 5 an example, the evaluation confidence score 554a is represented by the first graphical bar, and the evaluation confidence score 554b is represented by the second graphical bar, where the first bar has a higher height than the second bar because the confidence level associated with the evaluation confidence score 554a is higher than that associated with the evaluation confidence score 554b. The evaluation confidence scores 554a, 554b can be expressed in one or more ways (e.g., only numerically, only graphically, only textually, or a combination thereof).

[0059] Moreover, in some embodiments, the evaluation confidence scores 554a and 554b report the estimated accuracy of the corresponding authenticity evaluations 552a and 552b. For example, in the case where the modification 526a includes the verification data 228, the modification 526a can be evaluated as credible, and the evaluation can have a high evaluation confidence score 554a. However, although the modification 526b is evaluated as likely not to be authentic, due to the lack of verification data for the corresponding verification data 228 in the modification 526b, or because it has not been verified, the low evaluation of the authenticity of the modification 526b can still have a relatively high evaluation confidence score 554b.

[0060] In combination Figure 2 and Figure 3 and with reference Figure 4 , flowchart 460 begins with receiving, using the neural network 342, the modified content 220 / 320 generated based on the baseline digital content 222, including one or more modifications 226a and / or 226b made to the baseline digital content 222 (step 462). As Figure 2 shown, the modified content 220 can be received from the device 210d via the communication network 204 and the network communication link 206. Step 462 can be executed by the software code 240 / 340 of the system 230, executed by the hardware processor 234, and using the neural network 342.

[0061] Continuing with flowchart 460, at step 464, each of the one or more modifications 226a and / or 226b made to the baseline digital content 222 is evaluated for its authenticity using the neural network 342 to generate one or more authenticity evaluations 352a and / or 352b corresponding to the one or more modifications 226a and / or 226b made to the baseline digital content 222, respectively. For clarity of concept, the steps outlined in flowchart 460 will be further described with reference to a specific use case, in which the modified content 220 / 320 includes the modifications 226a and 226b to the baseline digital content 222. However, it is noted that in other use cases, the modified content 220 / 320 can include one of the modifications 226a or 226b, or more than the modifications 226a and 226b.

[0062] Step 464 can be performed in a variety of different ways by the software code 240 / 340 executed by the hardware processor 234 using the neural network 342. For example, with reference Figure 1 , Figure 2 and Figure 3, in an implementation manner of using the secure transaction ledger 101 / 201 to track the implementation of authorized modifications to the baseline digital content 122 / 222, step 464 can be performed by obtaining the authorized modification history of the baseline digital content 122 / 222 from the secure transaction ledger 101 / 20 and comparing the authorized modification history with the modifications 226a and 226b included in the modification content 120a / 120b / 220 / 320. It should be noted that, in some implementations, the authorized modification history of the baseline content 122 / 222 can be obtained by comparing the baseline content 122 / 222 with the modification content 120a, and then comparing the further modification content 120b with the modification content 120a and / or the baseline content 122.

[0063] For example, in one use case, the user 102b can be an authorized editor of the baseline digital content 122 / 222 and can be assigned a private key for this purpose, although the user 102c may not be authorized to modify the baseline digital content 122 / 222. In the said use case, whenever the user 102b who holds the private key modifies the baseline digital content 122 / 222, the secure transaction ledger 101 / 20 is updated through the edit data 121a. In some implementations, a user without a private key, such as the user 102c, may not be able to receive or edit the content 122 or the modification content 120a, thereby preventing premature disclosure or tampering. After production, the secure transaction ledger 101 can be finalized by a supervisor or through a revocation key held by the user 102a and / or the user 102b. The assessment of the authenticity of the modifications 226a and 226b presupposes the ability to distinguish between the modified and unmodified parts of the baseline digital content 122 / 222. For example, in combination with Figure 1 , Figure 2 , Figure 3 and Figure 4 and referring to Figure 5 , step 464 can include using the neural network 342 to distinguish the modification segments 580a, 580d, and 580e of the modification content 120a / 120b / 220 / 320 / 520 from the baseline segments 580b, 580c, and 580f. In some implementations, based on the intrinsic properties of the baseline digital content 122 / 222 / 522, such as on a per-segment basis, the verification data 224 of the baseline digital content 122 / 222 / 522 can be used for the said distinction.

[0064] The inherent properties of the baseline digital content 122 / 222 / 522 on which the verification data 224 is based can include the content contained in each segment of the baseline digital content 122 / 222 / 522 itself, as well as the specific characteristics of each segment. Referring to the baseline segments 580b, 580c, and 580f, each of these unmodified baseline segments can be identified based on, for example, the amount of data (in bytes) contained in each and / or its data format. Alternatively or additionally, the inherent properties of the baseline segments 580b, 580c, and 580f on which the verification data 224 is based can include metadata, where the metadata can be a device identifier (device ID) identifying the device 110a used to generate each of the baseline segments 580b, 580c, and 580f, a software application identifier (application ID) of the digital content production software used to generate the baseline segments 580b, 580c, and 580f, and / or an identifier of the user 102a of the device 110a. Additionally, in some embodiments, the inherent properties of the baseline segments 580b, 580c, and 580f on which the verification data 224 is based can include metadata, where the metadata identifies the production date of each of the baseline segments 580b, 580c, and 580f, their respective production times, and / or the location of the device 110a at the time of producing each of the baseline segments 580b, 580c, and 580f, for example based on data recorded by one or more orientation / location sensors of the device 110a ( Figure 1 the orientation / location sensors of the device 110a are not shown).

[0065] Any or all of the above inherent properties, including each of the baseline segments 580b, 580c, and 580f, can be hashed using any suitable cryptographic hash function, such as one from the Secure Hash Algorithm (SHA) hash function family (e.g., SHA-0, SHA-1, SHA-2, or SHA-3). Thus, the verification data 224 includes the hash values of each of the baseline segments 580b, 580c, and 580f, as well as some, all, or none of the additional inherent properties of the baseline segments 580b, 580c, and 580f. In some embodiments, the verification data 224 can be appended to the baseline digital content 122 / 222 / 522. For example, as is known in the art, the verification data 224 can be appended to the baseline digital content 122 / 222 / 522 as a "sidecar" as is known in the art. However, in other embodiments, it may be advantageous or desirable to embed the verification data 224 within the baseline digital content 122 / 222 / 522.

[0066] Then, by generating verification data for each segment and comparing the verification data with the verification data 224 of each segment, the modified segments 580a, 580d, and 580e can be distinguished from the unmodified baseline segments 580b, 580c, and 580f. For example, in an implementation where the verification data 224 contains the hash values of each segment of the baseline digital content 122 / 222 / 522, the verification data for each segment 580a - 580f can be generated by hashing each segment using the same cryptographic hash function used to generate the verification data 224. The verification data generated in this way for the baseline segments 580b, 580c, and 580f will match the verification data 224 of these segments, while the verification data generated for the modified segments 580a, 580d, and 580e will not match the verification data 224 of the modified segments 580a, 580d, and 580e, thereby identifying the modified segments 580a, 580d, and 580e as modified content.

[0067] The assessment of the authenticity of each of the modifications 226a / 526a and 226b / 526b can be performed in a manner similar to that described above for distinguishing the modified segments 580a, 580d, and 580e from the baseline segments 580b, 580c, and 580f. For example, referring to the authorized modification 226a / 526a made to the baseline digital content 122 / 222 / 522 by the authorized user 102b using the device 110b, the authenticity of the modification 226a / 526a can be evaluated based on the inherent properties of the modification 226a / 526a and using the verification data 228 of the modification 226a / 526a.

[0068] The inherent properties of the modification 226a / 526a on which the verification data 228 is based can include the content contained in the modification 226a / 526a itself, as well as the specific characteristics of the modification 226a / 526a, such as the data size and / or its data format of the modification 226a / 526a. Alternatively or additionally, the inherent properties of the modification 226a / 526a on which the verification data 228 is based can include metadata, where the metadata identifies the device ID of the device 110b used to generate the modification 226a / 526a, the application ID of the editing software used to generate the modification 226a / 526a, and / or the identity of the user 102b of the device 110b. Furthermore, in some implementations, the inherent properties of the modification 226a / 526a on which the verification data 228 is based can include metadata, where the metadata identifies the production date, production time, and / or the location of the device 110b at the time of production of the modification 226a / 526a, such as based on data recorded by one or more location / orientation sensors of the device 110b ( Figure 1 the orientation / location sensors of the device 110a are not shown in the figure).

[0069] Any or all of the above intrinsic properties, including the modification 226a / 526a, can be hashed using any suitable cryptographic hash function, such as one of the SHA hash function families described above. Thus, the verification data 228 includes the hash value of the modification 226a / 526a and some, all, or none of the additional intrinsic properties of the above modification 226a / 526a. In some embodiments, the verification data 228 can be attached to the modification 226a / 526a as an "attachment box". However, in other embodiments, it may be advantageous or desirable to embed the verification data 228 in the modification 226a / 526a. Embedding the verification data 228 in the modification 226a / 526a can advantageously avoid stripping the verification data 228 from the modification 226a / 526a.

[0070] The authenticity of the modification 226a / 526a can then be evaluated by using the neural network 342 to generate verification data for the modification 226a / 526a using one or more of the above intrinsic properties of the modification 226a / 526a and comparing the verification data with the verification data 228 of the modification 226a / 526a. For example, in an implementation where the verification data 228 includes the hash value of the modification 226a / 526a, the verification data for the modification 226a / 526a can be generated by hashing the modification 226a / 526a using the same cryptographic hash function used to generate the verification data 228. That is, the verification data for the modification 226a / 526a includes the hash value of one or more intrinsic properties of the modification 226a / 526a. The verification data generated in this way for the modification 226a / 526a will match the verification data 228 of the modification 226a / 526a, thereby identifying the modification 226a / 526a as likely authentic. In contrast, a modification 226b / 526b that lacks verification data corresponding to the authentic modification 226a / 526a will remain unverified during the evaluation process.

[0071] Step 464 causes the neural network 342 to generate an authenticity assessment 352a / 552a corresponding to the modification 226a / 526a and / or generate an authenticity assessment 352b / 552b corresponding to the modification 226b / 526b. As described above, the authenticity assessment 352a / 552a can identify the modification 226a / 526a as a genuine modification by an authorized user 102b to the baseline digital content 122 / 222 / 522. In contrast, the authenticity assessment 352b / 552b can identify the modification 226b / 526b as unverified and thus likely not authentic.

[0072] Continuing with flowchart 460, at step 466, a credibility assessment 270 / 370 / 570 of the modified content 120a / 120b / 220 / 320 / 520 is generated based on the credibility assessment 352a / 552a and / or the credibility assessment 352b / 552b. As described above, in some embodiments, the software code 240 / 340 may omit the scoring module 344. As Figure 3 shown, the neural network 342 is configured to directly output the credibility assessment 352a and / or the credibility assessment 352b to the assessment module 346. Thus, in some embodiments, the hardware processor 234 may execute the software code 240 / 340 and use the assessment module 346 to generate a credibility assessment 270 / 370 / 570 of the modified content 120a / 120b / 220 / 320 / 520 based on the credibility assessment 352a / 552a and / or the credibility assessment 352b / 552b.

[0073] However, in embodiments where the software code 240 / 340 includes the scoring module 344, the hardware processor 234 may execute the software code 240 / 340 to determine an assessment confidence score 354a / 554a for the segment 580a of the modified content 120a / 120b / 220 / 320 / 520, and / or determine an assessment confidence score 354b / 554b for the segments 580d and 580e of the modified content 120a / 120b / 220 / 320 / 520. As described above, the assessment confidence scores 354a / 554a and 354b / 554b report the estimated accuracy of the corresponding credibility assessments 352a / 552a and 352a / 552b. For example, as further discussed above, in the case where the modification 226a / 526a is evaluated as authentic based on the validation data 228, this evaluation may have a high assessment confidence score 354a / 554a. Additionally, although the modification 226b / 526b is evaluated as potentially inauthentic, the low assessment of the authenticity of the modification 226b / 526b may still have a relatively high assessment confidence score 354b / 554b due to the lack of validation data for the corresponding validation data 228 of the modification 226b / 526b or the modification 226b / 526b not being verified.

[0074] In embodiments where the assessment confidence score 354a / 554a and / or the assessment confidence score 354b / 554b is determined, the assessment module 346 may be used to generate a credibility assessment 270 / 370 of the modified content 120a / 120b / 220 / 320 / 520 based on the assessment confidence score 354a / 554a and / or the assessment confidence score 354b / 554b and the corresponding credibility assessment 352a / 552a and / or the credibility assessment 352b / 552b.

[0075] The method illustrated in flow chart 460 ends at step 468: Output the authenticity assessments 270 / 370 / 570 of the modified content 120a / 120b / 220 / 320 / 520 for presentation on a display. For example, as Figure 2 shown, in some embodiments, the authenticity assessments 270 / 370 / 570 of the modified content 120a / 120b / 220 / 320 / 520 can be transmitted via communication network 204 and network communication link 206 to device 210d for presentation on display 212. It is noted that in some embodiments, hardware processor 234 can execute software code 240 / 340 to perform steps 462, 464, 466, and 468 in an automated process where human intervention can be omitted.

[0076] Accordingly, the present application discloses a system and method for assessing the authenticity of modified content, which overcomes the drawbacks and deficiencies in the conventional art. By using a trained neural network to assess the authenticity of one or more modifications made to an original or "baseline" version of digital content. The present application discloses an ingenious verification solution that enables a detailed assessment of the modified content. In addition, by generating verification data for assessing the modified content based on the inherent attributes of one or more modifications made to the underlying baseline digital content, this solution conveniently utilizes the characteristics of each modification to assess its authenticity.

[0077] Based on the foregoing description, it is apparent that various techniques can be used to implement these concepts without departing from the scope of the concepts described in this application. In addition, although these concepts have been specifically described with reference to certain embodiments, those of ordinary skill in the art will recognize that changes can be made in form and detail without departing from the scope of these concepts. In this regard, the described embodiments are to be considered in all respects as illustrative and not restrictive. It should also be understood that this application is not limited to the specific embodiments described herein, but that many rearrangements, modifications, and substitutions are possible without departing from the scope of this disclosure.

Claims

1. A system, comprising: A computing platform including a hardware processor and a system memory; Software code stored in the system memory, the software code including a neural network trained to evaluate the authenticity of modified content generated based on baseline digital content, the modified content including one or more modifications made to the baseline digital content; The hardware processor is configured to execute the software code to: Receive the modified content using the neural network; Use the neural network to distinguish one or more modifications made to the baseline digital content from the baseline digital content; Use the neural network to evaluate the authenticity of each of one or more modifications made to the baseline digital content and distinguished from the baseline digital content, to generate one or more authenticity evaluations corresponding to the one or more modifications made to the baseline digital content and distinguished from the baseline digital content; Generate an authenticity evaluation of the modified content based on the one or more authenticity evaluations; And Output the authenticity evaluation for presentation on a display.

2. The system according to claim 1, wherein the neural network is configured to generate verification data for each of the one or more modifications based on one or more intrinsic attributes of each of the one or more modifications used by the neural network, so as to evaluate the authenticity of each of the one or more modifications made to the baseline digital content, and wherein the one or more intrinsic attributes include one or more of the following: the amount of data in each segment of the baseline digital content, the data format of each segment of the baseline digital content, the device identifier of each device used to generate each segment of the baseline digital content, the production date of each segment of the baseline digital content, or the location of the device at the time of production of each segment of the baseline digital content.

3. The system according to claim 2, wherein The verification data for each of the one or more modifications includes a hash value of one or more intrinsic attributes of each of the one or more modifications.

4. The system according to claim 1, wherein The authenticity evaluation includes a visualization graphically depicting the one or more authenticity evaluations.

5. The system according to claim 4, wherein, The visualization includes a heat map composed of the modified content based on the one or more authenticity evaluations.

6. The system according to claim 1, wherein The hardware processor is configured to further execute the software code to determine an evaluation confidence score for one or more segments of the modified content based on the one or more authenticity evaluations.

7. The system according to claim 6, wherein The authenticity evaluation includes a heat map of the modified content based on the evaluation confidence scores of one or more segments of the modified content.

8. The system according to claim 1, wherein, Each of the one or more segments of the modified content is segmented based on a timecode interval of the modified content.

9. The system according to claim 1, wherein, The modified content includes video, and wherein each of the one or more segments of the modified content includes one of a video scene, a video shot, or a video frame.

10. The system according to claim 1, wherein, The modified content includes audio, and wherein each of the one or more segments of the modified content includes an audio track.

11. The system according to claim 1, wherein, Before receiving the modified content, the system is configured to: Verify the authenticity of content creation or content modification; and Train a neural network using training data obtained from verified content.

12. The system according to claim 11, wherein, Verify authenticity to verify that the user created or modified the content.

13. The system according to claim 12, wherein, Verify authenticity using metadata identifying the device or application used to create or modify the content in the content.

14. A method used by a system including a computing platform, the computing platform having a hardware processor and a system memory storing software code, the software code including a neural network trained to evaluate the authenticity of modified content generated based on baseline digital content, the modified content including one or more modifications made to the baseline digital content, the method including: Receiving the modified content by software code executed by the hardware processor and using the neural network; Distinguishing, by software code executed by the hardware processor and using the neural network, one or more modifications made to the baseline digital content from the baseline digital content; Evaluating, by software code executed by the hardware processor and using the neural network, the authenticity of each of one or more modifications made to the baseline digital content and distinguished from the baseline digital content to generate one or more authenticity assessments respectively corresponding to the one or more modifications made to the baseline digital content and distinguished from the baseline digital content; Generating, by software code executed by the hardware processor, an authenticity assessment of the modified content based on the one or more authenticity assessments; And Outputting, by software code executed by the hardware processor, the authenticity assessment for presentation on a display.

15. The method according to claim 14, wherein, The neural network is configured to generate verification data for each of the one or more modifications based on one or more intrinsic attributes of each of the one or more modifications used by the neural network, thereby evaluating the authenticity of each of the one or more modifications made to the baseline digital content, and wherein the one or more intrinsic attributes include one or more of the following: the amount of data in each segment of the baseline digital content, the data format of each segment of the baseline digital content, the device identifier of each device used to produce each segment of the baseline digital content, the production date of each segment of the baseline digital content, or the location of the device at the time of production of each segment of the baseline digital content.

16. The method according to claim 15, wherein, The verification data for each of the one or more modifications includes a hash value of one or more intrinsic attributes of each of the one or more modifications.

17. The method according to claim 14, wherein, The authenticity assessment includes a visualization graphically depicting the one or more authenticity assessments.

18. The method according to claim 17, wherein, The visualization includes a heat map consisting of the modified content based on the one or more authenticity assessments.

19. The method according to claim 14, further including determining, by software code executed by the hardware processor, an evaluation confidence score for one or more segments of the modified content based on the one or more authenticity assessments.

20. The method according to claim 19, wherein The authenticity assessment includes a heat map of the modified content based on the evaluation confidence score of one or more segments of the modified content.

21. The method according to claim 14, wherein, Segment each of one or more segments of the modified content according to the time code interval of the modified content.

22. The method according to claim 14, wherein, The modified content includes video, and each of one or more segments of the modified content includes one of a video scene, a video shot, or a video frame.

23. The method according to claim 14, wherein, The modified content includes audio, and each of one or more segments of the modified content includes an audio track.

24. The method according to claim 14, wherein Before receiving the modified content, the method further includes: verifying the authenticity of content creation or content modification; and training a neural network using training data obtained from the verified content.

25. The method according to claim 24, wherein, Verifying authenticity verifies that a user created or modified the content.

26. The method according to claim 25, wherein, Verifying authenticity uses metadata identifying a device or application used to create or modify the content in the content.

Images