How to Detect Deepfake Videos

Deepfakes are not a new phenomenon and have been around for a while. However, technological advances have made them quite sophisticated. Recently, the deepfake video of Indian actor Rashmika Mandana emerged, and in the past, many celebrities have been targeted by such videos. We explain what deepfake videos are, how they are made and how to detect them:

How are deepfake videos made?

Deepfakes are created using a combination of encoder and decoder networks, often in the context of a generative adversarial network (GAN). The encoder network analyzes the source content (e.g., the original face) and extracts essential features and representations. These features are then passed to the decoder network, which generates new content, such as a manipulated face.

How to Detect Deepfake Videos

To detect deepfakes, you can look for visual and audio inconsistencies and other telltale signs.

Look for facial expression abnormalities

Check for unnatural facial expressions, mismatched lip sync, or blinking irregularities.

Listen to the audio carefully for discrepancies

When you have questions about a video, look for discrepancies in the audio. Listen for changes in tone or unnatural speech patterns.

Check the surroundings of the video

Look for visual distortions, blurring, or inconsistent lighting. Check whether the person in the video can be in that environment.

Context and content are the most important

Analyze whether the behaviour or statements align with the individual's known characteristics.

Check the source of the video

Confirm the legitimacy of the media from reliable sources. Is it a random YouTuber or social media account posting the video? If yes, look for other discrepancies.