CAVEAT – We want to thank Toddington International, a great company that teaches open source intelligence programs, and who we have trained with personally, for educating us on some ideas when it comes to helping internet users identify an online deepfake photo. For the purpose of this article, we are going to look at deepfakes specific to the online targeting of both teens and adults when it comes to catfishing, sextortion, and online sexual predation. It should also be noted that all the pictures in this article are deepfakes!
*** Updated 2024-01-17 ***
A deepfake (also known as synthetic media) is an artificial intelligence-based digitally forged image or video of a person that may or may not exist. Such images or videos can depict a person of any age, race, sexual orientation, or gender identity.
Given that we are licensed online investigators, we covertly investigate predatory talk groups in the deep web. We have seen discussions, by those who want to prey upon others for financial gain or sexual predation, talking about the use of deepfake technology as a part of a targeted pretext strategy. It is because of this fact, we believe that youth and even adults need to be aware of deepfakes; more importantly – how to identify a deepfake. Given that deepfake technology is constantly adapting and maturing, what we discuss in this article may become moot as artificial intelligence used to generate these images becomes more sophisticated.
According to Toddington International, there are 8 “tells” that can help in identifying a deepfake photo:
#1 – Have a close look at the ears (our favorite). Presently, deepfake technology has difficulty in creating ears that look realistic and are symmetrical – especially the earlobes which will differ from ear to ear.
#2 – Have a close look at any jewelry. Deepfake technology has a challenge in creating realistic-looking jewelry. It’s also common that jewelry, especially earnings, won’t match.
#3 – Have a close look at the glasses. Often, the features of the glasses will not be symmetrical with the frames and any of its hardware, such as the hinges on the frames.
#4 – Have a close look at the clothing. Often, the clothing will lack detail and will be mismatched especially when it comes to clothing texture
#5 – Have a close look at the background. Often, deepfake-generated images will use a blurred background effect, commonly known as bokeh, or items in the background will not look symmetrical and will appear warped.
#6 – Have a close look at the teeth, especially any of the back teeth that are viewable. Often the back teeth will be a different texture, size, and color when compared to other teeth.
#7 – Have a look at any facial hair. Often there will be viewable inconsistencies especially when it comes the symmetry of a mustache, beard, and eyebrows
#8 – Have a look at the skin which can sometimes look inconsistent throughout the picture and may even appear to be rubbery.
The above-noted information was specific to deepfake photographs, but another pathway that online predators are using – live real-time streaming video-based deepfake technology. Although youth may think they are streaming live with a 16-year-old female from California, it could be a 45-year-old male from South East Africa, using streaming-based real-time deepfake technology, combined with voice morphing software to make them look and sound like a teen.
Recently, Metaphysic, a company that specializes in creating AI technology to create your own hyperreal avatar, released a research paper to help identify a deepfake live-streaming video (1). In their research, Metaphysic found that current deepfake live streaming video technology has two weaknesses – if you have the person on the other end of the streaming video #1 turn their head to at least a 90-degree angle, and #2 quickly have them move their hand in front of their face, you will see clear distortion in both cases. If the other person doesn’t want to comply with your request – Danger, Danger, Danger! This safety strategy should be a standard operating protocol whenever streaming with anyone online who you do not know personally or are virtually meeting for the first time online.
As deepfake technology continues to develop, it is going to become harder and harder for those who use technology to identify a real person from a deepfake. Thankfully, today there are still discrepancies to help us to identify the difference if you know what to look for, thus the reason for this article.
Let’s Put The Above Noted Information Into Practice:
Now with your students, or as a family, here’s a great website for you to apply what you have learned in this article into practice – have fun! https://www.whichfaceisreal.com
Recently a company called “Deepware” and “aiornot” have now also provided a FREE online resource to help identify if a video may have used deepfake technology you can locate here – https://scanner.deepware.ai and https://www.aiornot.com
Here’s an actual case study that we conducted on a video where we identified several “artifacts” that lead us not to believe the authenticity of the video, and that it had likely been digitally altered https://thewhitehatter.ca/blog/how-deepfake-technology-ai-can-be-used-for-mis-dis-information-to-support-a-narrative-a-case-study/
Here’s a great FREE training resource from another online investigator Ben Brown, that we highly recommend to help us identify AI Generated Fake Images https://www.youtube.com/watch?v=h_YzuXQlIXI
The White Hatter
References
(1) https://metaphysic.ai/to-uncover-a-deepfake-video-call-ask-the-caller-to-turn-sideways/
