Deepfakes
What are they and how to spot them
Deepfakes – videos, images and audio
What they are and the best way to spot them
Deepfakes are created when deep machine learning is used to generate fake content – which may seem real. Sometimes, it’s just for fun – but often they’re deliberately misleading. Audio deepfakes can be easier to generate (and harder to spot) than video ones.
Deepfakes can be used as part of a Social Engineering attack. Examples include impersonating a client’s voice on a trade request confirmation call or featuring the exact likeness of a CEO stating untrue information.
Seeing is not always believing - how to spot deepfakes
Read our guide to help you recognize the warning signs and safeguard your online interactions.
Listen carefully. Badly made deepfakes use robotic voices, mispronunciation, poor lip-syncing, or no audio at all.
Facial features. Unnatural facial hair can identify a fake. Look out for extra facial moles or strange lip size and color. If you focus just on the mouth, you often can tell it looks "weird" which is enough to be wary of it.
Blinking. Researchers found that in some deepfake videos, people don’t blink like we do (or at all). Replicating how we blink is difficult. So, look for unnatural or slow blinking.
Lighting. Unusual lighting, strange skin tones, stains, and glasses with too much glare, are all signs you might be looking at a fake. Also, watch out for eyes and eyebrows where you would not expect them to be, or with too much shadow.
Context is key. It’s important to consider where the content came from and whether the source is a trustworthy one. Unexpected contact from someone on an unusual platform could be the work of a cybercriminal.
