Unmasking Deepfakes: Tips For Identifying Manipulated Videos

Highly realistic computer-generated voices can convincingly deceive people, while sophisticated masks crafted from social media images can compromise facial recognition security like Face ID. These advancements, enabled by AI-driven deepfake technology, are no longer the stuff of science fiction but are now exploited by malicious actors targeting common people.

After the recent Rashmika Mandanna deepfake video incident, discussions about this contentious technology are escalating, with social media users increasingly seeking ways to identify such videos.

Identifying deepfake videos can be challenging, especially as the technology continues to evolve. However, there are several indicating signs that can help you spot a deepfake content. Here are some things to look for:

Visual inconsistencies

According to the Massachusetts Institute of Technology, focus on facial details while analysing a deepfake video. High-quality deepfakes primarily involve facial transformations, so observe the cheeks and forehead for unnatural smoothness or excessive wrinkles, MIT said, asking users further to  ensure the skin's age matches the hair and eyes. Users are also asked to pay attention to the eyes and eyebrows, looking for accurate shadows and a proper representation of natural physics in the scene. Things like change in glare on glasses according to angle changes as a person moves can also come in handy. Then of course, the authenticity of facial hair, including moustaches, sideburns, or beards should also be checked. 

Audio inconsistencies

Checking the lip sync: This is among the most basic checks to find out if a clip if digitally manipulated. Does the person's mouth move correctly when they speak? If the lips don't match the words, it could be a fake video.

Listen to the audio: Experts advise people to check if the sound is clear and natural. In case the audio sounds muffled or distorted, it could be fake.

Pay attention to the background noise: Does the background sound match the scene? If the background noise is inconsistent or unnatural, it could be fake.

Context identities

When assessing a video's credibility, observe if the depicted actions align with the person. Verify the source's credibility and check for fact-checking by reputable organisations.

Deepfake detection tools

As deepfake technology has progressed, so have the methods and technologies developed for its detection. The tech industry has been developing several deepfake detection tools. Here are some of the most useful tools:

1. Sentinel

2. Intel's Real-Time Deepfake Detector

3. WeVerify

4. Microsoft's Video Authenticator Tool

5. Deepfake Detection Using Phoneme-Viseme Mismatches