“Deepfakes”: Identifying and Detecting Them

Deepfake technology has rapidly evolved and has been employed for various purposes, some harmless and entertaining, while others malicious and harmful. Deepfakes are multimedia content, including images or videos, in which a person’s face or body is digitally altered to make them appear as a different individual.

Initially, deepfake technology was mostly used for comedic purposes, creating amusing content like superimposing the faces of famous Malayalam actors onto iconic characters from “The Godfather.” While such videos were often easy to identify as fake, the quality of video and audio synchronization improved, leading to viral content on platforms like Instagram.

However, deepfake technology is now being misused to malign high-profile individuals, including politicians and movie actors, with the intent of damaging their reputation, especially before elections or a movie release.

To identify deepfake content, there are several telltale signs to watch out for:

1. Pay close attention to the start of the video as the person’s face may not change immediately, and the deepfake effect might only become apparent after a certain point.
2. Observe the person’s facial expressions throughout the video, looking for irregular changes during conversations or actions.
3. Check for lip sync issues, as there may be minor audio/video synchronization problems in deepfake videos.
4. Look for variations in body posture that may not align with a real person’s behavior.
5. Always verify the source of the video by searching for the content on reputable search engine platforms before jumping to conclusions.

There is a pressing need for both government agencies and technology companies to collaborate on developing cross-platform detection tools to combat deepfake videos. Ensuring that these tools are accessible to the general public and implementing regulations for deepfake creator apps to include watermarks for easy identification are crucial steps in curbing the spread of fake news and deepfake content.

_{Repurposed article originally published in the Deccan Herald}