Technology

10 ways to spot a LinkedIn deepfake, like that message in your inbox

Amrutha PagadApril 1, 2022 | 14:37 IST

Imagine you are scrolling through your Twitter or Facebook feed and you come across a video of Prime Minister Narendra Modi in his 8 pm address-to-the-nation-style, announcing that he’s stepping down from his role. What would your reaction be?

You are going to be shocked! You are likely to forward the video to a group on WhatsApp. But there’s one thing you are unlikely to suspect – is it real? Is the video real?

You may want to check again; there’s a strong likelihood that it’s fake – more precisely, it's a ‘deepfake’.

What is a deepfake? Deepfake is essentially a synthetic media which is manipulated to make someone real say fake things or even make someone completely unreal using artificial intelligence. Here is Elon Musk as a WWE fighter. 

via GIPHY

Why are we talking about deepfakes? While the PM Modi scenario has not yet come to pass, it’s not difficult for someone to spread fake news such as that. It’s happening elsewhere in the world in much graver scenarios and it’s happening even in the most mundane of things – like the corporate messages in your LinkedIn inbox.

Deepfake during war: Amidst the chaos, destruction and heartbreak of the Ukraine war, a video of Ukrainian President Volodymyr Zelenskyy went viral, where he is seen seemingly ordering the Ukrainian forces to lay down their arms and surrender to the invading Russian forces.

Except, the very real-looking Zelennskyy video is unreal. Imagine the consequences of such a video in such a situation.

In another video, there was Russian President Vladimir Putin declaring victory in Ukraine. That too was not real.

These videos were the result of deepfakes. 

Deepfake for corporates: But the technology is not just being used to make politicians say the weirdest things or make Tom Cruise do a magic trick (remember that?). It is also being used for the most mundane of things like everyday marketing.

Renee DiResta and her colleague Josh Goldstein, researchers at the Stanford Internet Observatory, recently found out that LinkedIn is being flooded with deepfake profiles pitching products for companies.

Perhaps if you were to open your LinkedIn inbox right now, you may find a message from someone new and perhaps someone deepfake.

DiResta found out about the deepfakes on LinkedIn when she was messaged by one Keenan Ramsey. It seemed like a normal message at first with Ramsey introducing herself as a member of a group that DiResta belonged to on the job-searching social media platform. Then she pitched for a software subscription. It all seemed very real at first glance.

Deepfakes become marketing tools on LinkedIn. Illustration: Geetanjali/DailyO

But then DiResta’s veteran misinformation-hunting eye caught something very amiss in the profile picture of the person. And after some digging, the researcher came to the conclusion that it was in fact a deepfake, that the profile and the profile picture were not that of a real person, but a bot – a sales bot in this case. Following this, DiResta and Goldstein dug deeper and found that there were more than 1000+ such deepfake profiles, mostly selling stuff.

Their research found that marketing companies were creating these fake avatars for sales pitches and in some cases offering these fake profiles for a sum to other corporates.

What's in it for corporates? It is part of a post-pandemic strategy for digital marketing which enables companies to pitch directly to their customers. Except, using fake profiles is a cheaper method of doing so, instead of hiring that number of sales rep. When someone responds to the bot positively, then they are redirected to a real person, a sales rep who closes the deal.

But creating such fake profiles is against LinkedIn’s guidelines. The company is said to have taken action after DiResta and Goldstein’s research.

But how will you ensure that what you have come across is not a deepfake? Here are 10 ways to detect a deepfake:

via GIPHY

Deepfakes can be in any form – video, audio or photo.

1. In the case of a video, if it’s a poorly done deepfake, it can be easily recognised. Look for unnatural eye movements or the eyes not blinking at all. In the case of the Zelenskyy video, the unnatural voice and the unmoving body were a dead giveaway.

2. Look for a mismatch of expressions or emotions when one image is stitched over the other. In another case, there is likely to be a bad lip-sync or facial movement. Sometimes, facial features may seem misaligned.

3. Most of the videos are also of a very low resolution to make detection harder.

4. Unnatural skin tone can also be noted along with discolouration. For example in the Zelenskyy video, the colour of his neck and face seemed out of match.

5. In the case of images, you can look for unnatural eye placement. DiResta found that the LinkedIn deepfakes had one thing in common – the eyes in the profile pictures were perfectly aligned to the middle of the image.

via GIPHY

6. You can also look for weird or unnatural-looking hair in pictures. Awkwardly sticking out hair or hair that seems to blur in the background is usually one which is cropped from the original.

7. DiResta found that in some deepfake LinkedIn profiles, the background was unusually blurred or that the background wasn’t discernible.

8. Zooming in on such pictures can also help detect discrepancies or watching a video in slow mode.

9. Reverse image search can also help look up on Google whether the video or image already exists.

10. For videos, there’s a website called Deepfake-o-meter which looks up whether a video is a deepfake or not for you. But this method could take some time to determine.

So, go check your inboxes – on LinkedIn or even on mail. Do you think the sales messages are from deepfake bots?

Last updated: April 01, 2022 | 14:37
IN THIS STORY
Read more!
Recommended Stories