A deep dive into deepfakes

Viktor Underwood, CEO at Quickchannel, warns deepfakes have the power to disrupt and damage trust in broadcasting

By Contributor

Published: July 29, 2021

Viktor Underwood, CEO at Quickchannel, warns deepfakes have the power to disrupt and damage trust in broadcasting

target=_blank title=Share on LinkedIn class=share-linkedin>

The internet has already got a taste of the disruptive power of deep fake technology, i.e. manipulated video clips that reproduce fabricated images and audio that appear to be real. Some deep fakes' have already become viral hits on social media, whether it's a clip of a comedian morphing into a movie star on a talk show or a global leader using some colourful language to describe a political opponent.

While the former is a great example of how deep fakes can be used for entertainment, the latter offers a more chilling view of how this sophisticated AI-powered technology can be used for more nefarious purposes, such as seminating disinformation and undermining trust in governments.

The prospects become all the more concerning if one were to consider deep fakes' power to take fake news' to the next level it wouldn't be out of place to expect popular news broadcast stations to get the deep fake treatment, whereby falsified newscasts are produced and shared across social media. When viewed through this lens, it becomes clear why combating the rise of deep fakes is set to be a defining trend in the broadcasting industry.

Seeing is believingBack in June 2020, Sensity, a deep fake detection technology firm, identified over 49,000 deep fake videos online a staggering increase of more than 330% since July 2019. A study conducted by Nanyang Technological University found that one in three Singaporeans who said they were aware of deep fakes believed they had at one point circulated content on social media which they later found out to be a hoax.

What's most disconcerting about these findings is not so much the rate at which deep fake videos are multiplying, but rather how frequently viewers will believe what they see and how easily they will share it with their network. With the technology evolving at a rapid pace, it won't be long before deep fakes will be able to fool even the most alert viewers.

data-src=https://www.tvbeurope.com/wp-content/uploads/2021/07/deepfake-queen-726x408.jpeg alt= width=449 height=252 data-srcset=https://www.tvbeurope.com/wp-content/uploads/2021/07/deepfake-queen-726x408.jpeg 726w, https://www.tvbeurope.com/wp-content/uploads/2021/07/deepfake-queen-353x199.jpeg 353w, https://www.tvbeurope.com/wp-content/uploads/2021/07/deepfake-queen-768x432.jpeg 768w, https://www.tvbeurope.com/wp-content/uploads/2021/07/deepfake-queen-241x136.jpeg 241w, https://www.tvbeurope.com/wp-content/uploads/2021/07/deepfake-queen.jpeg 1280w data-sizes=(max-width: 449px) 100vw, 449px />Channel 4 broadcast an alternative Christmas message delivered by a deepfake Queen in 2020It's also worth noting that these challenges raise an important issue about trust and integrity when it comes to broadcasting content. Authenticity and transparency are becoming increasingly important values for the average individual on the internet, which means that preventive measures need to be put in place if we are to tackle the spread of fake content effectively.

Detection leads to deflectionThe good news is that measures are already being implemented to mitigate the damage that deep fakes are able to inflict. For example, Twitter has put policies in place that outlaw deep fake technology. As a result of these policies, the social media platform is able to tag any deep fake videos that have yet to be removed. YouTube on the other hand put a ban on deep fake videos related to the 2020 US Census, as well as election and voting procedures.

Detection and filtering software also has a fundamental role to play. Microsoft announced the availability of a tool to spot deep fakes back in September 2020, whereby the software analyses photos and videos to give a confidence score about whether the material is likely to have been artificially created. More recently, Facebook scientists revealed they can create software that reverse engineers deep fake images to understand how they were made and where they originated.

But perhaps the most effective preventive measure against fake content is actually to build viewer awareness. Growing volumes of fake content and continuous advances in technology are only set to make the task more difficult, so training people to spot telltale signs might actually yield more results than actively trying to contain the spread. Bad lip-syncing, lack of blinking, as well as unnatural eye movement and facial expressions are all key signs in helping to identify and flag deep fakes.

Apply a personal touch to your contentAt this point, it is clear that any form of video content has the potential to be deep faked', with varying degrees of efficacy. This presents a challenge considering that, following the events of 2020, people are only going to increasingly rely on secure, high-quality video streaming, while businesses capitalise on web-based media streaming to get their message heard.

This puts added pressure on video content tech companies, who are taking preventive measures against deep fakes by implementing consistent communication structures, centralised monitoring and effective detection practices. However, what some businesses might not appreciate yet is the value of investing in personalised video content as a means for tackling deep fakes. After all, in a world where widely available content can be altered with relative ease, personalising your broadcast content might just be the most effective way to tackle this issue.

One of the most common pieces of advice given today is to make sure that the content we cons

LINK:	https://www.tvbeurope.com/production-post/a-deep-dive-into-deepfakes...
	See more stories from tvb

Most recent headlines