Deepfake

AI-generated video, images, or audio of a person, often used by fraudsters to attempt gaining access to a locked account or service.

For example, a fraudster might use software to make a video of a real person's face say or do things they never did, then use it to try to pass a verification check or fool a colleague on a call.

Synthetic media produced by generative models, typically face-swap or voice-cloning techniques trained on real footage of the target, to impersonate a specific identity in a way that passes casual visual or audio inspection.

Deepfakes come in several common types:

• Face-swap deepfakes place one person's face onto another person's video.
• Voice cloning recreates a person's voice from a short audio sample.
• Synthetic avatars render an entirely fabricated person, which can run in real time on a video call.

A deepfake attack against identity verification usually has two parts. The attacker creates the fake, then delivers it, either by showing it to the camera or by injecting it into the app.

Deepfakes sit next to terms that are easy to mix up:

• Deepfake detection is the technology that tells synthetic media apart from real captures.
• Injection attack is a delivery method that attackers often use for deepfakes.
• Synthetic identity fraud builds an identity from real and invented data, and a deepfake can supply the face for it.

For more, see Deepfakes 101: Examples and deepfake fraud prevention tips, The 3 Most Common Types of Facial Deepfakes Explained, and the webinar Deepfakes in 2026.