Quantcast
Connect with us

Deepfakes: Here’s how to detect fake online political videos with one simple tell

Published

on

A new form of misinformation is poised to spread through online communities as the 2018 midterm election campaigns heat up. Called “deepfakes” after the pseudonymous online account that popularized the technique – which may have chosen its name because the process uses a technical method called “deep learning” – these fake videos look very realistic.

ADVERTISEMENT

So far, people have used deepfake videos in pornography and satire to make it appear that famous people are doing things they wouldn’t normally. But it’s almost certain deepfakes will appear during the campaign season, purporting to depict candidates saying things or going places the real candidate wouldn’t.

It’s Barack Obama – or is it?

Because these techniques are so new, people are having trouble telling the difference between real videos and the deepfake videos. My work, with my colleague Ming-Ching Chang and our Ph.D. student Yuezun Li, has found a way to reliably tell real videos from deepfake videos. It’s not a permanent solution, because technology will improve. But it’s a start, and offers hope that computers will be able to help people tell truth from fiction.

What’s a ‘deepfake,’ anyway?

Making a deepfake video is a lot like translating between languages. Services like Google Translate use machine learning – computer analysis of tens of thousands of texts in multiple languages – to detect word-use patterns that they use to create the translation.

Deepfake algorithms work the same way: They use a type of machine learning system called a deep neural network to examine the facial movements of one person. Then they synthesize images of another person’s face making analogous movements. Doing so effectively creates a video of the target person appearing to do or say the things the source person did.

ADVERTISEMENT

How deepfake videos are made.

Before they can work properly, deep neural networks need a lot of source information, such as photos of the persons being the source or target of impersonation. The more images used to train a deepfake algorithm, the more realistic the digital impersonation will be.

Detecting blinking

There are still flaws in this new type of algorithm. One of them has to do with how the simulated faces blink – or don’t. Healthy adult humans blink somewhere between every 2 and 10 seconds, and a single blink takes between one-tenth and four-tenths of a second. That’s what would be normal to see in a video of a person talking. But it’s not what happens in many deepfake videos.

A real person blinks while talking.
A simulated face doesn’t blink the way a real person does.

When a deepfake algorithm is trained on face images of a person, it’s dependent on the photos that are available on the internet that can be used as training data. Even for people who are photographed often, few images are available online showing their eyes closed. Not only are photos like that rare – because people’s eyes are open most of the time – but photographers don’t usually publish images where the main subjects’ eyes are shut.

ADVERTISEMENT

Without training images of people blinking, deepfake algorithms are less likely to create faces that blink normally. When we calculate the overall rate of blinking, and compares that with the natural range, we found that characters in deepfake videos blink a lot less frequent in comparison with real people. Our research uses machine learning to examine eye opening and closing in videos.

This gives us an inspiration to detect deepfake videos. Subsequently, we develop a method to detect when the person in the video blinks. To be more specific, it scans each frame of a video in question, detects the faces in it and then locates the eyes automatically. It then utilizes another deep neural network to determine if the detected eye is open or close, using the eye’ appearance, geometric features and movement.

We know that our work is taking advantage of a flaw in the sort of data available to train deepfake algorithms. To avoid falling prey to a similar flaw, we have trained our system on a large library of images of both open and closed eyes. This method seems to work well, and as a result, we’ve achieved an over 95 percent detection rate.

ADVERTISEMENT

This isn’t the final word on detecting deepfakes, of course. The technology is improving rapidly, and the competition between generating and detecting fake videos is analogous to a chess game. In particular, blinking can be added to deepfake videos by including face images with closed eyes or using video sequences for training. People who want to confuse the public will get better at making false videos – and we and others in the technology community will need to continue to find ways to detect them.The Conversation

Siwei Lyu, Associate Professor of Computer Science; Director, Computer Vision and Machine Learning Lab, University at Albany, State University of New York

This article was originally published on The Conversation. Read the original article.


Report typos and corrections to: [email protected].
READ COMMENTS - JOIN THE DISCUSSION
Continue Reading

2020 Election

‘I don’t care’: Watch Kamala Harris shut down Chris Hayes for asking a dumb question about Trump

Published

on

Sen. Kamala Harris shut down MSNBC anchor Chris Hayes during a post-debate interview on Tuesday evening.

Hayes questioned Harris about her call for Twitter to follow their terms of service and kick President Donald Trump off of the platform.

"Do you think he puts people’s lives in danger when he targets them in tweets?" Hayes asked.

"Absolutely," Harris replied.

"Do you think he knows that?" Hayes asked.

"Does it matter?" Harris replied.

"The fact is he did it. The fact is that he is irresponsible, he is erratic," she explained. "He is like a 2-year-old with a machine gun."

Continue Reading

2020 Election

Democrats blast Trump and demand his impeachment at CNN debate

Published

on

Democratic White House hopefuls united in searing condemnation of Donald Trump during their fourth debate Tuesday, saying the president has broken the law, abused his power, and deserves to be impeached.

From the opening moments, most of the dozen candidates on stage launched fierce broadsides against Trump over the Ukrainian scandal at the heart of the impeachment inquiry.

"The impeachment must go forward," said Senator Elizabeth Warren, who is neck and neck with former vice president Joe Biden at the head of the 2020 nominations race.

"Impeachment is the way that we establish that this man will not be permitted to break the law over and over without consequences," she thundered.

Continue Reading
 

2020 Election

Here are 3 winners and 4 losers from the CNN/NYT Democratic presidential primary debate

Published

on

Twelve Democrats took to the stage Tuesday night for yet another debate in the party's 2020 president primary hosted by CNN and the New York Times.

After only ten candidates qualified for the previous debate, an additional two — Hawaii Rep. Tulsi Gabbard and wealthy donor and former hedge fund manager Tom Steyer — made it to the stage this round for an even more crowded event.

The candidates discussed a range of important policy issues, but since the format was a debate, and they're all competing for the same nomination, it is ultimately most critical who won and who lost the night. Here are three winners and four losers — necessarily a subjective assessment, of course — from the debate:

Continue Reading
 
 
Help Raw Story Uncover Injustice. Join Raw Story Investigates for $1 and go ad-free.
close-image