Take a look of this video of Donald Trump posted on Belgian social media:
Most likely, you noticed a few odd things about it, right? The above video is an example of a low-quality deepfake. Most viewers wouldn’t hesitate to call it out because the video looks fake.
Now, take a look at this video of Barack Obama:
This video looks much more real. Obama looks, and sounds, like the real Obama. This is the type of deepfake content that poses the greatest threat.
Is that a deepfake?
The Pentagon’s DARPA agency has already spent tens of millions on a media forensics research program, but while they develop an adequate deepfake detection system, we must rely on our naked eye to spot deepfakes.
Here are some basic ways to spot a deepfake:
Curious lack of blinking: Deepfake creators need source material — images and video of the person being impersonated. Often the source material doesn’t provide enough blinking imagery. While adults blink between every 2 and 10 seconds, the blinking rate for deepfakes is generally less frequent. Also, a person usually blinks more often when talking, a physiological reality not necessarily reflected in the majority of deepfakes.
Watch the mouth: After the eyes, the mouth is the next critical area that is most likely to give away a deepfake. If audio and lip-syncing are not synchronized you might be looking at a fake. Also, keep an eye on the inside of the mouth. Many deepfake creators have not yet figured out how to make teeth, tongue, and the mouth detailed enough.
Verify the source first: Was the video you are watching shared by a credible source? If not, perform a Google search using the video’s main keywords. The results may help you uncover a deepfake. Be especially wary of video content published by individual users and clickbait or deeply partisan outlets.
Unnatural physiological signals: In manipulated videos, it is common for the subject’s breathing to be incongruent with speech. Also, a static facial expression and lack of natural pauses may be telltale signs that something is amiss. If you have watched real video footage of the subject multiple times, your intuition may tell you something is not right.
Protecting your reputation
Although complete protection against deepfakes or other types of online disinformation is nearly impossible, there are certain preemptive measures we can take.
Approach all online communications strategically: If you are known for maintaining a poised personal brand online, the sudden appearance of a deepfake showing implausible behavior will be received with skepticism. Additionally, your followers will likely be the first to report the negative content and show their support. Finally, if you frequently post video footage, your followers will be familiar with your communication style, making it easier for them to spot the ruse.
Monitor what is being said about you: You need to be the first one to find out if there is an attempt to damage your reputation online. If you have an online communications team tracking relevant mentions of you, you will be able to discover and diffuse negative sentiments online before they escalate.
Analyze mentions: Someone who takes their time to deepfake you has most likely already displayed antagonistic behavior towards you online. By monitoring your mentions you can take measures against these users by flagging them and even taking legal action before they employ a more thorough plot to harm your reputation.
Have official records of public appearances: Ask your communications team to film and retain recordings of your public appearances so you always have authentic media footage to dispel false footage that may spread online.
Add digital watermarking to your video content: Truepic uses blockchain technology to store the unique information of your images and video. If you use Truepic’s camera feature, every photo or video you take will have a watermark with a copy of the image it saves, so viewers can compare them to ensure the version they’re seeing hasn’t been altered. This type of forensics can help you debunk deepfakes.
Ignore it: If the deepfake has low reach and impact it is best to do nothing. If nobody is sharing the malicious content it will expediently fall right into internet oblivion. Giving it attention can fire back and make it more visible.
We cannot control what is published on the internet. However, we can and should, control our online presence and educate our employees or other members of our organization on how to detect and combat harmful deepfakes. For more information on the implications of deepfakes read this article.
DON’T GET CAUGHT BY SURPRISE
We’ll send you our monthly newsletter, The Angry Mob, detailing stories of leaders on the wrong side of the public’s pitchfork.