Could the job applicant be a fake? Impostor uses deepfake to steal identity to apply for remote work, FBI: can recognize when speaking

​Produced by Big Data Digest

Author: Caleb

Online work is gradually officially entering our lives.

According to statistics from data company Emsi Burning Glass, at the beginning of 2020, of the 163,000 job openings in New York City, only 6,700 allowed online work, accounting for 4% of the total number of positions; but by last year In December, among the 243,000 recruitment positions in the city, the number of online jobs increased to 25,800, accounting for 10.6% of the total number of positions.

But as the saying goes, people’s hearts are separated by the belly, not to mention that there is an extra layer of screens between them. No one can guarantee that the Neso incident will not happen again.

No, according to recent FBI investigations, more and more people are stealing other people's personal identity information and using deepfakes to apply for remote jobs. It is precisely because deepfake technology uses artificial intelligence or machine learning to generate content such as images, videos, or audios that it is difficult to distinguish it from real materials.

For example, according to foreign media reports, a scammer has been working remotely with the help of deepfake in an attempt to gain access to IT jobs and access their customer or financial data, as well as corporate IT data and professional information, etc.

Coincidentally, according to reports from other companies, some job seekers’ personal information belongs to another person entirely.

Videos and voices are all faked. Is it difficult to see through them?

On June 28, the FBI Cybercrime Complaint Center stated in a public consultation that the number of complaints about people being impersonated for job applications has increased recently. Scammers have been using deepfake technology and personally identifiable information stolen from victims to trick employers into hiring them for remote or work-from-home jobs.

These jobs involve IT, computer programming, database and software-related positions.

The identity disguise implemented by deepfake is often difficult to distinguish between authenticity and fraud, and scammers can steal valuable information from within the company. details, and conduct other identity fraud schemes.

Even according to federal law enforcement agencies, during the investigation, they also discovered that the interviewer’s voice used in the online interview had also been forged by deepfake.

It is unclear how many people have successfully joined the company using this method and how many of them have been reported.

It is not completely impossible to identify deepfake. The FBI says there is a way for employers to detect deepfakes. In their daily work, employees still need to video chat with their employers from time to time to report on work progress, etc. When they speak, that is when their flaws are revealed.

"In front of the camera, if the movements and lip movements of the person being interviewed cannot be fully synchronized with the voice", "such as a sudden cough, sneeze or other auditory behavior and visual presentation "The content is inconsistent", that is often the key to seeing through deepfake.

But it’s not that easy to see through deepfake, especially if you don’t pay special attention. According to a recent report by Carnegie Mellon University researchers, the accuracy of artificial intelligence designed to detect doctored videos could range from 30% to 97%. Humans have ways of detecting fake videos, especially once they Trained to see certain visual glitches, such as abnormal shadows or problematic skin texture.

How deepfake is used depends on the person

Whether deepfake is used well or not, is it used on the "edge" , but also depends on the person.

In 2021, former Beatle Paul McCartney used deepfake technology to "return to youth". In the MV with Beck, the audience sees a young, flexible McCartney, with no trace of time left on him.

Musk was also deepfake last month.

In this video of TED Chairman "Chris Anderson" interviewing Musk, Musk said that a cryptocurrency trading platform called BitVex was created by himself and promised that within three months, any crypto Currency deposits earn 30% daily returns.

#deepfake also does good things sometimes.

Recently, in a case that reopened the investigation in the Netherlands, in order to find out the truth about 13-year-old Sedar Soares who was shot and killed while throwing snowballs with his friends, with the permission of Sedar Soares’ family Next, the police used deepfake to create a video of Sedar Soares to collect clues from the public and find witnesses. In the minute-long video, images of Soares' life appear as he greets the camera and picks up a football. The day after the video was released, Rotterdam police spokesperson Lillian van Duijvenbode said: "The fact that we have received dozens of tips is very positive."

Again, the value orientation of technology depends on the people who use the technology.

Exercise your deepfake recognition ability

As an AI deep learning tool, deepfake can replace the face of a person in a picture. Technology that takes someone else's picture to create a very realistic "fake" video or picture, hence the name "face swapping".

The researchers used the autoencoder neural architecture to make this idea a reality. The basic idea is very simple: train a set of encoders and corresponding decoding neural networks for each face. When coding, the picture of the first person was used. When decoding, a second human decoder is used.

In order to help people better identify deepfake scams, researchers also collected 100,000 deepfakes hosted in Kaggle public competitions videos and 19,154 real videos, a series of neural networks were trained to detect deepfakes.

In general, there is no completely absolute method to help identification, but there are several points worth noting in deepfake:

• Pay attention to the face: High-end DeepFake manipulations are almost all based on the face;
• Pay attention to the cheeks and forehead: Does the skin appear too smooth or too wrinkled? Does skin aging match the aging of hair and eyes? Deepfakes tend to be jarring in some ways;
• Pay attention to the eyes and eyebrows: are the shadows in the right place? Deepfakes often fail to replicate scenes with natural physics completely correctly;
• Watch out for glasses: Is there any glare? Is there too much glare? Does the angle of glare change when a person moves? Once again, DeepFakes often don't fully represent the natural physics of lighting;
• Pay attention to facial hair, or lack thereof: does facial hair look realistic? Deepfake may add or remove beards, sideburns or beards, but such behavior often brings a sense of disobedience;
• Pay attention to moles on the face: Do moles look real;
• Pay attention to blinking: Whether the person blinks enough or too much;
• Pay attention to the size and color of the lips: whether the size and color match the rest of the person's face.

These eight questions can help people check and identify deepfake.

High-quality deepfake is not easy to identify, but through practice, people can establish an intuition based on the authenticity of deepfake. Just trust the intuition and use appropriate auxiliary verification. .

If you are still worried, you can also take a simple test on this website: https://detectfakes.media.mit.edu/​

The above is the detailed content of Could the job applicant be a fake? Impostor uses deepfake to steal identity to apply for remote work, FBI: can recognize when speaking. For more information, please follow other related articles on the PHP Chinese website!