Deepfake: Funny Pranks or Disruptive Fraud?

How It Works

The word “deepfake” is a mash-up of the words “deep learning,” which is associated with machine learning and artificial intelligence (AI), and an old-fashioned reference to something being “fake.”

Fraudsters start by taking advantage of an abundance of content available of a specific individual for free from online news outlets and social media, including consumer-generated video and photos. The fraudster manipulates that content into a new image or voice recording with the subject saying or doing something that they never did or that is out of character for them.

The process itself is very technical: Deepfakes are generated by neural networks, using artificial intelligence and machine learning algorithms to imitate real humans. However, an advanced degree is not required to create deepfakes thanks to readily available software that enables people with basic computer skills to become deepfake creators. Quality of the final product may reflect skill level, but may also improve over time due to machine learning, not necessarily human learning.

Who Gets Hurt

Those most often victimized are politicians, celebrities and high-level business executives, due in part to the availability of images, videos and audio recordings of them on the internet. Sometimes the goal of the fraudster is a funny prank that drives social media engagement. But other times the goal is to spread misinformation, ruin reputations or inflict financial harm.

Don’t assume you won’t be impacted by deepfake if you are not a Hollywood or Washington, D.C., socialite. An internet search on deepfakes highlights a wide array of fraud. For example, workers at corporations have been duped into taking inappropriate actions when instructed by a deepfake of their CEO instructing them to do so, sometimes resulting in significant costs and job losses. At a minimum, you may be misinformed about an important topic or person due to deepfakes.

How To Detect Deepfakes

Deepfake images, audio and video can be so well done that it is very difficult to determine a fake. Because deepfakes are synthesized using high tech, detecting them with high tech and algorithms is key. In general, here are a few things to be on the lookout for:

  • Badly synced sound and video
  • Blurriness where the face meets the neck and hair
  • Box-like shapes or other cropped effects around the mouth, eyes and neck
  • Changes in the background and/or lighting
  • Face discolorations Irregular blinking
  • or no blinking Lower-quality sections in the same video
  • Movements that aren’t natural

Important Legal Disclosures & Information

Read a summary of privacy rights for California residents which outlines the types of information we collect, and how and why we use that information.