top of page

Deepfakes

Frame 7.jpg

A deepfake is a fake image, audio (‘voice skins’ or ‘voice clones’), text, or video generated by specific artificial intelligence/machine learning algorithms, namely deep learning. Deepfakes are made to be believable and realistic, even though the events captured in the media never happened. Deep fakes can come in many forms, whether it be pornographic content or altering politicians' interviews to spread misinformation. While researchers have identified key ways to identify weaknesses in this generated, fake content, it is still not easy to detect (Sample, 2020; What the Heck Is a Deepfake? | UVA Information Security, n.d.).

​

Statistics

  • The AI firm Deeptrace found 15,000 deepfake videos online in September 2019, a near doubling over nine months. A staggering 96% were pornographic and 99% of those mapped faces from female celebrities onto porn stars (Sample, 2020).

  • In 2020, researchers found that over 100,000 computer-generated fake pornographic images of women were created without their consent or knowledge (What the Heck Is a Deepfake? | UVA Information Security, n.d.).

  • Approximately 90-95% of deepfake videos since 2018 are based on non-consensual pornography (What the Heck Is a Deepfake? | UVA Information Security, n.d.).

  • 70 percent of people have said that they are not confident that they cannot tell the difference between a real and deepfake voice (“2024 Deepfakes Guide and Statistics,” n.d.).

How Adolescents can be Harmed

Scams/financial fraud:

  • Deepfakes can be personalized videos that show a relative or friend asking for money to help them out. This can be accompanied by a link or way in which to transfer money, and subsequently, can result in scamming people out of their money (What the Heck Is a Deepfake? | UVA Information Security, n.d.).

  • An example of this happened when the YouTuber Mr Beast appeared to offer his followers new iPhones for $2, but it wasn’t actually him (MrBeast and BBC Stars Used in Deepfake Scam Videos, 2023).

Fake pornographic videos and photos:

  • Non-consensual pornography represents a large majority of the computer generated, fake content on the internet. For example, individuals' faces have been placed on to pornographic images or videos, and subsequently sent around on the internet and social media. Francesca Mani was a 15-year-old girl when she, along with her friends, were targeted in a series of fake, sexually-explicit images that were generated by their classmates on computers. The New York Time’s podcast, The Daily, reported on Francesca’s situation and subsequent calls for change in policy in the episode, “Real Teenagers, Fake Nudes” (Chow, 2024; Real Teenagers, Fake Nudes: The Rise of Deepfakes in American Schools - The New York Times, n.d.; What the Heck Is a Deepfake? | UVA Information Security, n.d.).

Blackmail:

  • Deepfakes can consist of images, videos, text and audio that are used to blackmail individuals. People can create media of others and use it as leverage to force those people into doing tasks, such as transferring money, otherwise the individuals run the risk of damage to their reputation and exposure of inappropriate or explicit content (albeit fake).

Spread of Misinformation:

  • For example, fake videos of important world leaders, speaking on sensitive topics and making false claims, have been generated and have the potential to incite violence or even in the most harsh cases, war (What the Heck Is a Deepfake? | UVA Information Security, n.d.).

  • This is an example of the Tom Cruise deepfakes whereby he was depicted telling a story about him talking to someone in Russia. The video is a deepfake and is not representative of Tom Cruise.​

Practical Guidance

Do not send or share videos or photos with people that you do not know.

  • In December 2019, a journalist researching deepfake pornography and the technology behind it, paid a fake content creator to insert her face into pornography. Instead of needing to submit a lot of photos to create the video, the journalist found that a 15 second video from an Instagram story, containing 450 individual frames, had all of the frames needed to create the video. The journalist sent in a 13 second video of herself talking to a camera and the link to a Pornhub video where she wanted her face switched (What the Heck Is a Deepfake? | UVA Information Security, n.d.).

​​

Digital footprint:

  • Understand that data and photo contents that are shared with others on the internet are permanent.

​

Audience selectors and privacy controls:

  • Make your social media accounts private and only allow those you know to follow you

  • Only post to friend groups or close friends, not publicly or to people that you do not know personally.

  • Block people or accounts that seem suspicious and unfamiliar.

​​

Fact check with credible sources:

  • Double check information that is seen on the internet especially in videos or audios

How to Identify  a Deepfake

  • Audio

    • An individual should look for the following signs when trying to determine if an audio is fake:

      • Choppy sentences

      • Changing of tone in speech

      • Phrasing (would that person say it in that way?)

      • Context of message (does it relate to current events or previous issues that the person has spoken about?)

  • Video:

    • Blinking (or lack thereof), movements that are not natural or mismatched

    • Non symmetrical accessories such as earrings

    • Facial features lacking definition, especially if the rest of the image/background is not obscured

    • Lower-quality sections throughout the same video or changes in the background and/or lighting

    • Different skin tones around the face, specifically towards the edge of their face.

    • Double chins, double eyebrows, or double edges to the face

    • Whether the face gets blurry when it is partially obscured by a hand or another object

    • Harsh cut-off lines, shapes, and effects around the mouth, eyes, and neck that appear jagged.

    • Contextual clues, does the background match the foreground?

    • The lip synching could be poorly timed

    • Smaller details, such as hair, are hard for deepfakes to create well, especially when there are strands visible on the edge

    • Badly rendered jewelry and teeth can also be a giveaway, as can strange lighting effects, such as inconsistent illumination and reflections on the iris (Sample, 2020).

What to do if a Deepfake was Made about you or your Friend

  • Report any suspected deepfakes immediately

  • Report inappropriate content and abuse on social media platforms (i.e. Facebook, Twitter, Instagram, etc.) using the platforms’ reporting procedures (What the Heck Is a Deepfake? | UVA Information Security, n.d.).

Sources

© 2035 by Joop. Powered and secured by Wix

bottom of page