What Is Proactive Media Forgery Prevention?

In the previous article, we discussed deepfake and cheapfake detection. Such forgery verification technology is used to detect forgery after the media has been tampered with or edited. Academic circles and related industries are concerned that even new detection technologies and methods of the future will not prevent media forgery completely. Thus, you might wonder whether there is any technology that will completely prevent media forgery. Check out the following proactive media forgery prevention technology.

Pre-Blocking Principle

First, let's take a look at media forgery prevention technology that is used internationally. As classified below, image forgery verification is divided into active and passive verifications.

Deepfake and cheapfake detection fall under the passive verification category. Numerous techniques have been introduced to detect image manipulation in the passive area. However, it wasn't easy to detect them perfectly because forgery technology had advanced at a much faster rate. On the other hand, active verification can verify and prevent forgery 100%.

Image Forgery →Active → Digital watermarking Image Forgery →Active → Digital Signature Image Forgery → Passive → Image Splicing Image Forgery → Passive → Image Cloning Image Forgery → Passive → Image Retouching
Source : Blind Approach for Digital Image Forgery Detection, international Journal of Computer Applications vol. 179-no.10, Jan 2018 (https://www.ijcaonline.org/archives/volume179/number10/28839-2018916108)

Then how is media forgery in the active area detected 100%? Read on to learn what kind of technologies help prevent forgery.

JPEG Privacy & Security

Images taken using smartphone cameras are created following the international JPEG standard. Images created through such digital shooting devices are distributed through multiple platforms. In this process, images can be forged, thus causing distortion of facts.

The JPEG International Standardization Organization (ISO) has established and applied privacy & security standards to prevent digital image forgery. For digital signatures in the JPEG privacy & security area, a hash value of the image and certain metadata* is extracted and made into an encryption key. The signed data is included in a certain metadata area called the JPEG Universal Metadata Box Format (JUMBF).
* Metadata refers to information in the JPEG file, such as the device, time, and place of generating, taking, and storing the image.

Therefore, every image contains the hash value of the original image. If the metadata or image is modified, the hash value changes, so you can easily check whether the image was modified by checking whether the hash value has changed. Team9 applied this technology for the first time in the world and is using it to prevent digital forgery.

Signature-based method Support JPEG-1 (ISO/IEC 10918), JPEG XT(ISO/IEC 18477), and JPEG2000 (ISO/IEC 15444) Embedding module Verification module Embedding module JPEG image SOL APP1 (Exif) SOS image data EOI image data APP1 (Exif) -> Hash function -> signature 8AEFB06C 426E07A0A671A1E2 488B4858D694A730 1. Open image 2. Collect metadata for integrety checking 3. Generate signature 4. Embed signature & metadata 5. Protected image Signature P&S metadata JPEG image SOI APP1 (JUMBF BOX) SOS image data EOI Verification mudule JPEG image SOI APP1 (Exif) APP11 (JUMBF Box) SOS Image date EOI Signature P&S Metadata Image data APP1 (Exif) Hash function -> 8AEFB06C 426E07A0A671A1E2 488B4858D694A730 1. Open protected image 2. Read signature & metadata 3. Generate signature 4. Compare both signature 5. Verify Signature ↔ signature compare
Materials related to JPEG privacy & security (Source: Professor Oh-jin Kwon of College of Electronics and Information Engineering, Sejong University)

Multi-point Focus

Developed by Team9, Multi-point Focus (MpF) distinguishes a three-dimensional object from a two-dimensional one. First, the three-dimensional image is split horizontally and vertically. Focusing the camera on one of the split segments will result in different levels of sharpness among segments. If you direct the camera focus to another segment, the sharpness will change depending on the distrance in focus. The team got the idea from such changes in sharpness. They took photos of several three-dimensional and two-dimensional objects and taught AI how to distinguish whether an object is three-dimensional or two-dimensional.

Three-dimension 1 2 3 4 5 6 7 8 9 Two-dimension 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
Samsung SDS, “The Method of Taking and Reading Images and Device for It,” cited from domestic/overseas patent applications (April 2020)
Photo taken focusing the center Photo taken focusing the background
Samsung SDS, DoFNet: Depth of Field Difference Learning for Detecting Image Forgery, Cited ACCV 2020

Time of Flight (ToF)

Time of Flight (ToF) is one of the light detection and ranging (LiDAR) sensors included in the terminal. The sensor measures the time it takes for a camera signal to reach the subject and then return back within the range. Team9 assumed that the ToF of a three-dimensional object is different from that of a two-dimensional object. In other words, three-dimensional objects have different ToF depending on the surface curvature, whereas the ToF of a two-dimensional object has little difference.

ToF (Time of Flight) - One of LIDAR sensors (Light Detection And Ranging) - The time is measured when the signal flights from the smartphone sensor that reaches the subject and returns back within the range . Used in precise calculation of distance between the camera and the subject
Samsung SDS, Team Nine Show Case data (April 2020)

Based on these measured data, Team9 has secured the technology to determine whether a photographed object is two or three-dimensional.

Analyze whether the cup is 3 or 2 dimensional
Samsung SDS Team Nine show case data (April 2020)

Artifact Detection

Moiré patterns appear differently in photoshoots that took the real object, paper, monitor, etc. Artifact analyzes these characteristics and learns whether the photograph is a real object, or a forged image on a monitor or printed on paper. Refer to the previous article for more information: What Are Cheapfakes (Shallowfakes)?

Real, Print, Screen photos of baby license
Samsung SDS, Team Nine internal test data (April 2021)

Adobe CAI

Adobe Photoshop is a representative image editing program. Adobe created the Content Authenticity Initiative (CAI) to validate all types of digital content, including photos and videos. In particular, the initiative developed an open-source that can add a layer of verifiable trust, leading a safer system (C2PA: https://c2pa.org/) that can verify digital content sources through collaboration and operation between companies.

With CAI technology, one can check the details of how an image was shot, edited, posted, or shared via social media. As seen below, the user can check various information regarding shooting, editing and posting by clicking the ⓘ icon in the upper right corner of an image.

Verity Beta Upload Image FAQ Learn more SNL_20201115_100701_M.jpg Content credentials ? These assets are in chronological order, starting with the latest version. Select one to explore itsd own content credentials. i FILE NAME SNL_20201115_100701_M.jpg i FILE NAME IMG_20201115_100701.jpg Compare credentials Contenet producer ? PRODUCED BY Sara Lewkowicz PRODUCED WITH Ps Photoshop PRODUCED ON 11/17/2020, 12:29 AM Edits and activity ? Color and adjustments Changes in color or tone Effects and styles Changes to a layer's appearance, like filters or layer styles Resizing and moving Changes to size, orientation, direction, or position Providers ? PRODUCER IDENTIFIED BY Adobe SIGNED BY Adobe Copyright c 2021 Adobe. All rights reserved. Privacy Terms of use Contact us
Adobe CAI, How it works ’20 (https://contentauthenticity.org/how-it-works)
(Source : https://www.youtube.com/watch?v=Xd6vtHMlse4)

The Adobe CAI community includes major IT companies such as Microsoft and Twitter, as well as manufacturers such as ARM and Nikon, and media companies, including the BBC. The CAI is expected to gather greater influence as the technology manages content history generated by the digital camera sensor, and users can retrieve the information on the cloud. Team Nine is also reviewing whether to include such features in the solution later.

Media Forgery Prevention Platform by Team9

If you take a picture with Team9's mobile forgery prevention application, it will be sent to the server immediately.

CYNETHIC Gallery @2021 SAMSUNGSDS rightsd reserved Launch Cyethic App Camera Hi, Kwansik Yoon. Welcome Take a picture and upload it. Camera Gallery Picture Click camera tab Take photo Click gallery tab choose photo to upload check uploaded photos in picture tab
Team9's Media Forgery Prevention Application

Photos taken in real-time with this dedicated application go through the following steps: server transfer > blockchain linkage > storage. The application verifies the authenticity of photos via real-time analysis, and the result can be viewed in a PDF report file.

CYETHIC contact Us sign up Login to Recharge Direct If you don't have an account You can Sign Up here! ID adam75 Password ....... Login Remeber me Forgot you Password? Request a reset link. English CYETHIC Home Task request Task in progress Task complete Hello, Kwansik Yoon Have a nice day at work! Recent Closed Tickerts Fail Warning Close table Ticket No Recipient Agent Last Updated 10998 navermi... minki hong 2021/05/222221/13:35 10123 창현 박 JaeHyeon Lee 2021/04/21 13:53 10096 박창현님 changhyun... 2021/04/16 13:25 Verification Analytics May 2021 close 61 Review 84 PASS WARNING FAIL N/A Closed Verification Chart Weekly PASS 33% WARNING 33% FAIL 33% Review Verificationn Chart Weekly PASS 17% WARNING 83% FAIL 0% Submit Timeline Today SF Ticket No 11043 재 Ticket No 11019 재 Ticket No11018 KY Ticket No11017 NT Ticket No 11042 Yesterday KY Ticket No 11031 IT Ticket No 11034 재 Ticket No 11035 재 Ticket No 11036 IS Ticket No 11037 재 Ticket No 11038 IE Ticket No 11033 CYETHIC Home Task request Task in progress Task complete Task request *Requied Infotmation Field REQUEST Request Mode* Select request mode Email Mobile Kakao Talk Talk Email teamnine@samsung.com First Name * First Name Last Name * Last Name Message Take your picture! Please download and run the CYETHIC app from the link below. You may refer to the guide while using the app. 171/500 Close Send Task in progress Total 537 Create 1 Open 1 Upload 13 Review 522 Ticket No Please enter a search term Verification Pass Warning Fail N/A Last Updated Today 1Week 1Month Reset Search Total 537 Status change *You cannot change the status in the Total tab. Status change Ticket No verification Recipient Agent Last Updated Upload Upload Review Review Upload Upload Review 11043 11042 11041 11040 11039 11038 11037 WARNING WARNING WARNING PASS WARNING PASS WARNING sds fifteen Note Ten test yj android two sds fourteen 재현 이 ios sds_thirsty chun seungbum changhyun park chun seungbum chun seungbum JaeHyeon Lee chun seungbum 2021/05/27 10:36 2021/05/27 08:07 2021/05/26 16:03 2021/05/26 15:38 2021/05/26 22:16 2021/05/26 14:37 2021/05/26 14:52

Forged media data are still circulating on the Internet as well as originals or copies of photos and videos, wreaking damage to individuals. Team9 developed a forgery prevention technology to block media forgery and verify media authenticity, thus preventing distortion of facts.

Next time, we'll look at cases of blocking deepfake and cheapfake and the related business areas.

※ This article was written based on objective research outcomes and facts, which were available on the date of writing this article, but the article may not represent the views of the company.