The De Lima video that made the rounds of online sites approaches the magnitude of deep fakes. Screengrab from @artsamaniego on Twitter
Culture Spotlight

Deep fakes bring fake news to a whole new level—and here's how it can target you

Identity theft is one area of concern in deep fakes. Facial scans are part of biometric identification—the Face ID lock of the iPhone for instance—and there are fears that a deep fake face can eventually fool a facial scan. Or in the case of the Leila De Lima video, lots of gullible netizens. 
Dominic Ligot | Aug 26 2019

Earlier this month a video of Sen. Leila De Lima made the rounds in social media, stirring public opinion about her situation, political affiliations, and motivations. I won’t go into detail about that video here (you might still find samples online) except on an interesting point—the video has started conversations about “deep fakes” and if they have finally arrived to the Philippines.

“Deep fakes” are computer generated images and videos created through a process called “deep learning” where a computer learns and generalizes patterns from images. Through complex information processing the computer is able to generate new “fake” images. (Hence the term). Deep learning itself is part of a broader term called machine learning that covers any process where computers learn from past data.

You may also like:

 
 
 
 
 
 
 
 
 
 
 
 
 

Wala ka sa lola ko #TheGeneralsDaughter

A post shared by Angel Locsin (@therealangellocsin) on

Although image processing and facial recognition research is not really new, the term “deep fake” has only become popular in the last 18 to 24 months when videos of popular Hollywood actors started to surface, their faces superimposed on other bodies with a surprising level of accuracy and realism. In 2017, researchers from the University of Washington were able to produce a photo-realistic visage of former US President Barack Obama, with the synthetic Obama’s head movements cued to any speech pattern. Since that time numerous videos and applications have surfaced all referencing similar tropes of image alterations:

  • FaceApp: Developed by Russian company Wireless Lab, the software generates filters that can transform any face into re-aged, re-gendered, re-styled, and smiling versions of the original.
  • FaceSwap: Apps that allow two people in an image or video to exchange faces.
  • DeepNude: Released by an anonymous programmer, the application takes any person’s image (usually a woman wearing a swimsuit) and transforms it into a completely nude person. This application has been banned on forums and social media recently on ethical grounds.

Going back to the De Lima footage, closer analysis suggests it to be the product of very sharp splicing which, while not a deep fake per se, still required creative and technical talent to achieve. Deep fake or not, the video’s context and intention appeared to put the subject under a different perspective and re-ordered and re-spun the subject’s words at significant odds with the original footage. It is on this note that social media and any form of information processing such as deep fakes starts to tread toward ethical dilemmas.

 

Reel and real

Identity theft is one area of concern in deep fakes. Facial scans are part of biometric identification—the Face ID lock of the iPhone for instance—and there are fears that a deep fake face can eventually fool a facial scan. Less onerous but more damaging is the potential for deep fake slander—someone impersonating you can say something damaging about you or anyone’s reputation which creates criminal and civil liabilities and paves the way for disinformation campaigns.

Someone impersonating you can say something damaging about you or anyone’s reputation which creates criminal and civil liabilities and paves the way for disinformation campaigns.

A less intuitive but glaring issue is the concept of data ownership i.e. a person’s visage. If anyone can grab a photo and create a digital version of you and make it say anything, does the original person have rights to enforce on the media featuring their new fake self? How about performers and actors who are faked into endorsing products or are fake acting in new scenes? Are they able to collect fees and royalties for their fake performance?

On the topic of performances, a current legitimate use for deep fakery is in movies. In the Marvel movie Captain America: Civil War, Robert Downey Jr. was de-aged in a flashback scene with Tony Stark’s parents, while in Star Wars: Rogue One, the visages of Carrie Fisher and Peter Cushing were digitally superimposed on Ingvild Deila and Guy Henry respectively to reprise the characters of Leia and Tarkin. In the movie The Social Network, the Winklevoss twins were played by Josh Pence and Armie Hammer but it was the latter’s face that was used on both.

Armie Hammer plays the Winklevoss twins in The Social Network.

It’s an interesting study of ethical contrasts in the use of a technology: deep fakes might be acceptable in movies, encouraged even where we are expected to suspend disbelief. But they are potentially criminally damaging in real life where reputations can be ruined.

Presumably, opportunities for creative expression and parody lie somewhere in between, like judging Will Ferrell’s portrayal of President Bush more hilarious when Bush’s face was deep faked onto it, or Jordan Peele’s more recent impersonation of Obama serving as a stern reminder of the risks.

Be wary of not just videos, but any politically-slanted, hard-sell, and emotionally-charged information you find online.

On the legislative front, Senator Vicente Sotto III opened the 18th Philippine Congress pushing Senate Bill No. 9, the “Anti-Fake News Bill.” This criminalizes and penalizes creators and publishers of fake news and disinformation. Media and human rights groups are criticizing the bill as restrictive of freedom of expression and online discourse, not to mention redundant given existing laws on libel, slander, and cybercrime exist. But it’s difficult to predict how perception will change if political and personal reputations start getting attacked by deep fakes eventually.

Meanwhile, just as machine learning research gave birth to deep fakes, a similar wave of study is looking at quickly detecting them, notably a study by US-based Defense Advanced Research Projects Agency (DARPA) and Facebook. As the arms race builds on both sides, the best defense the average citizen has against disinformation is a critical mind. Be wary of not just videos, but any politically-slanted, hard-sell, and emotionally-charged information you find online. In the era of social media, everyone can be a social journalist, which then implies that everyone should be a fact-checking social journalist before they share anything. 

“Truth is stranger than fiction, because fiction is obliged to stick to possibilities. Truth isn’t.”

Mark Twain might rethink his words today.

 

Dominic Ligot is a data analyst, software developer, entrepreneur and technologist. He is a founding board member of the Analytics Association of the Philippines where he is an active advocate for data literacy and data ethics. He is the founder of CirroLytix a machine learning research company. He previously held executive roles in IT and banking that included roles in governance, risk management, fraud, surveillance, and cybersecurity.

 

References:

1. Sotto’s Anti-Fake News Bill (https://news.abs-cbn.com/news/07/14/19/sotto-files-bill-vs-fake-news-disinformation)

2. HRW slams Anti-Fake News Bill (https://coconuts.co/manila/news/human-rights-watch-slams-senator-sottos-anti-fake-news-bill/)

3. Winklevoss Twins in The Social Network (https://www.theguardian.com/film/2010/oct/17/social-network-twins)

4. Rogue One Carrie Fisher (https://www.radiotimes.com/news/2017-03-30/get-a-closer-look-at-how-rogue-one-filmmakers-digitally-recreated-carrie-fisher/)

5. Deep Fake: Jordan Peele Obama (https://www.youtube.com/watch?v=bE1KWpoX9Hk)

6. Deep Fake: Will Ferrell Bush (https://www.youtube.com/watch?time_continue=2&v=bkP8-LcG-IE)

7. DARPA research in Deep Fake detection (https://www.technologyreview.com/s/611726/the-defense-department-has-produced-the-first-tools-for-catching-deepfakes/)

8. IPhone’s Face ID, How Secure (https://www.wired.com/story/iphone-x-faceid-security/)

9. Face Swap apps (https://beebom.com/best-face-swap-apps/)

10. Face App (https://www.faceapp.com/)

11. Risk and fall of DeepNude (https://www.vox.com/2019/6/27/18761639/ai-deepfake-deepnude-app-nude-women-porn)

12. University of Washington Obama Synthetic (https://www.bbc.com/news/av/technology-40598465/fake-obama-created-using-ai-tool-to-make-phoney-speeches)

13. What is a Deep Fake (https://fortune.com/2018/09/11/deep-fakes-obama-video/)

14. De Lima Video Splice (https://twitter.com/artsamaniego/status/1162181444602454016)