Researchers from Digimarc Corporation (NASDAQ: DMRC), inventor of the Digimarc Platform for digital identification and detection, will present details of a system for mitigating the problem of Deepfake news videos using digital watermarking at Electronic Imaging 2020 in Burlingame, CA, on Tuesday, January 28, 2020.  

The Digimarc research paper describes how watermarking technology can be used to prevent proliferation of Deepfake news. In the proposed system, digital watermarks are embedded in audio and video tracks of video clips of trusted news sources at the time the videos are captured or before they are distributed. The watermarks are detected at social media networks’ portals, nodes and back ends. The embedded watermark imparts a unique identifier to the video that links it to a blockchain. The watermarks also allow video source tracking, integrity verification and alteration localization. The watermark detectors can be standalone software applications, or they can be integrated with other applications. The watermark detectors are used to perform three main tasks: (1) alert the internet user when they watch an inauthentic news video, so that they may discard it, (2) prevent a Deepfake news video from propagating through the network, and (3) perform forensic analysis to help track and remove Deepfake news video postings. The paper includes proof of concept (POC) simulation results and will be available after the conference presentation on Tuesday.

“The Deepfake problem poses potentially catastrophic risks,” said Bruce Davis, CEO of Digimarc, who recently participated in an industry summit focused on content authenticity. “Digimarc’s Platform can help. There is much work to be done by industry and the research community to modernize media in ways that support continuing trust in what we see and hear via digital media that informs our decision-making and quality of life.”

Growing concerns challenge current laws
Concerns about the authenticity of news (text, audio, and video) distributed over the internet have reached an all-time high. In the past, people trusted news that came from reputable newspapers and trustworthy Radio/TV stations, but this trust is eroding for news distributed on the internet. The internet has enabled a non-linear media distribution model that does not guarantee the authenticity of news. Internet users can digitally alter news of authentic sources and re-distribute them through social media networks (e.g. YouTube, Facebook, Twitter, etc.) as if they were originals coming from legitimate sources. Usually, the alteration is done in three different ways. The first is known as face-swap, in which the original face in the video is replaced with another face. The second is known as lip-sync, in which the speaker’s voice is replaced by the voice of an impersonator. The third type is known as puppet-master, in which the person in the video is animated to do a desired action.

The news authenticity problem is exacerbated with the advent of deep learning technology. New powerful video creation software tools have recently been developed using deep learning and are now available on the internet for free, eliminating the need for the talents and specialized software typically found in the movie industry. A novice user can use an ordinary PC and leverage these tools to quickly alter the looks, the speech, or the actions of the people filmed in any video and generate fake videos that look convincingly real. The generated fake videos are commonly known as Deepfakes and their pervasiveness on the internet has doubled in the nine months period from December 2018 to July 2019 according to cyber-security company Deeptrace.

This rapid increase in the number of Deepfakes is alarming and their use could be detrimental to society. They have been used extensively for pornography and, to a much lesser extent, for cyberbullying celebrities, mocking renowned politicians, and robbing financial institutions. There is a growing concern that their harmful use could substantially increase, potentially influence elections, otherwise threaten security, and continue to diminish public trust in news media.

Congress and many states are introducing new legislation and policies to criminalize malicious Deepfakes. Non-profit organizations are running national campaigns to educate the public on how to deal with the danger of Deepfakes. These legislative actions and educational efforts help, but they are not adequate by themselves. It is imperative to develop advanced technical solutions to detect and prevent Deepfakes from penetrating social media networks.

System Architecture
The Digimarc research describes a system for detecting Deepfake news videos that use robust audio and image watermark and blockchain technologies.  The video and its metadata are stored for use in video forensic analysis in social media networks and other distribution platforms. Proof of concept simulations of the main parts of the system were performed. Preliminary results are encouraging, indicating that digital watermarking technology can be used successfully to link the video to its original copy and to the metadata stored in a blockchain network. They also show that the watermark embedded in the video can be detected after applying tools to create Deepfakes. The system can be generalized to include puppet-master Deepfakes and types of video other than news.

The paper will be available following the conference on the Digimarc website and on the Electronic Imaging Conference open access library. Other notable presentations by Digimarc researchers at the Electronic Imaging Conference include color imaging, Signal Rich™ art, detector optimization and digital watermarking for plastics sorting.

Digimarc Platform for Media Identification
Mitigation of Deepfake threats is an application of Digimarc’s highly effective auto-identification and data capture Platform, supported by a growing number of suppliers and application developers. Digimarc will be presenting three other papers at Electronic Imaging Conference describing elements of the Digimarc Platform, continuing it history of innovation and engagement with the academic community through publications and conference participation. To watch a brief video on how the Digimarc Platform works in retail applications, click here. For potential recycling uses, click here.  For more general information, visit digimarc.com and follow us @digimarc to learn more about The Barcode of Everything^®.

SOURCE Digimarc Corporation