Deepfakes: A Present Danger to Cybersecurity by@valentineenedah
410 reads

Deepfakes: A Present Danger to Cybersecurity

tldt arrow
Read on Terminal Reader

Too Long; Didn't Read

Deepfake is a video of a person who has been digitally manipulated in such that they appear to be someone else. Deepfakes can be employed maliciously to disseminate erroneous information. It is the 21st century’s answer to Photoshopping. It uses a subset of machine learning called Deep learning to make images/videos of fake events. The biggest achievement of the project was the subject's eyebrows not moving when the audio ceased playing. Previously, the deep fake was easier to see because it was unnatural when the subject continued to move while the subject was silent.

People Mentioned

Mention Thumbnail
Mention Thumbnail

Company Mentioned

Mention Thumbnail
featured image - Deepfakes: A Present Danger to Cybersecurity
Valentine Enedah HackerNoon profile picture

@valentineenedah

Valentine Enedah

I'm an active advocate for Technology.

react to story with heart

Technology is growing exponentially. Since the 1960s and 1970s, the speed and power of computers have typically doubled every one and a half to two years.🚀

But that’s not what we are talking about today!

Before I begin, I would love to ask a question⁉

Have you seen a video of Mark Zuckerberg boasting that he has complete control of stolen data from billions of individuals?

Or a video of  Sylvester Stallone(Rocky) inserted into Macaulay Culkin's role in Home Alone?😂

Or a video of an imitation of Morgan Freeman that looks so real?😮

Or the popular Kendrick Lamar music video that features him morphing into O.J. Simpson, Kanye West, Jussie Smollett, Will Smith, Kobe Bryant, and Nipsey Hussle.🔥🔥

If yes, then you have seen a “Deepfake”.

What is a Deepfake?

A Deepfake is a video of a person who has been digitally manipulated in such that they appear to be someone else. Deepfakes can be employed maliciously to disseminate erroneous information.

It is the 21st century’s answer to Photoshopping. It uses a subset of machine learning called Deep learning to make images/videos of fake events, hence the name “Deepfake”.

Its usage is increasing at an exponential rate, and it can pose a threat to cybersecurity.

image
Credit:  Sudeep Tanwar

Origin of Deepfake

A program was created in 1997 by Christoph Bregler, Michele Covell, and Malcolm Slaney to automate what some film studios could do with video rewriting.

On the basis of audio output, the program could also create fresh facial animations. This program was based on earlier work that recognized faces, created audio from text, and created the first realistic model of lips in three dimensions. Motion tracking eventually advanced as well as facial animations. Face2Face from the Technical University of Munich and the Synthesizing Obama Project from the University of Washington were both published as recently as 2016 and 2017.

Without audio, Face2Face projects an actor's mouth onto the target, but earlier research handled the synthesis of human voices. Credit:

The biggest achievement of the project was the subject's eyebrows not moving when the audio ceased playing. Previously, the deep fake was easier to see because it was unnatural when the subject's brows continued to move while the subject was silent.

The project's most astounding aspect was how long it took to compute a 66-second video using an NVIDIA TitanX and Core i7-5820. This process only takes a few hours on standard hardware.

A present danger to cybersecurity

Indeed, there are good applications of deepfakes such as improving medicine and education in general by creating learning tools.

It can be used in medicine to provide bogus patients whose data can be used for research purposes. This would safeguard patient information and encourage autonomy while offering researchers relevant data.

It is also quite popular in the gaming industry, AI-generated graphics can elevate the speed of game production.

Deepfakes in the gaming industry

Deepfakes in the gaming industry

Now, Deepfakes provide a cybersecurity risk to businesses since they might make phishing and Business email compromise (BEC) attacks more effective, facilitate identity fraud, and distort company reputations enough to cause an unjustifiable decline in share value.

Some criminals possess the required skills to create and use convincing deepfakes. Crime as a service is predicted to continue to expand and improve “in parallel with current technologies, leading to the automation of crimes such as hacking and adversarial machine learning and deepfakes," according to Europol.

Deepfake threats are classified into four main categories:

  1. sociocultural (sparking social instability and mass polarization)
  2. legal (distorting electronic evidence)
  3. personal (harassment and bullying, non-consensual pornography and internet child exploitation)
  4. standard cybersecurity (extortion, fraud and manipulating financial markets).

Deepfake photographs on forged passports make them tough to spot. These could then be used to support a variety of additional crimes, such as human trafficking, identity theft, illegal immigration, and travel by terrorists.

Extortion may be practiced through false allegations of shameful or unlawful behavior. If the bait includes a video or voice recording of a reliable acquaintance, phishing could advance to a new level. BEC assaults might be accompanied by a video message and voice that sound exactly like a notable figure of the organization, such as the CEO. Market manipulation still poses the most serious threat.

Scary right?💀

How do we protect ourselves against Deepfakes?

  1. Start by focusing on the cheeks and forehead. The skin could seem too wrinkled or too smooth.
  2. Is the skin's aging identical to that of the hair and eyes? deepfakes frequently have incongruent dimensions.
  3. Look at the eyebrows and the eyes. You might see the shadows where you would expect them to be. deepfakes frequently fall short of accurately capturing the scene's natural physics.
  4. There can be a glare, so pay attention to the glasses. As the individual moves, pay attention to how the angle of the glare changes because deepfakes frequently fall short of accurately simulating the actual physics of illumination.
  5. Pay attention to the presence or absence of facial hair. Is this facial hair realistic-looking? A moustache, sideburns, or beard could be added or removed by deepfakes. Facelifts with deepfakes frequently fall short of being completely natural.
  6. Pay attention to moles on the face.
  7. Be mindful of blinking. Is the person blinking too much or not enough?
  8. Pay attention to the lips' size and color because they can not be in proportion to the rest of the person's face.

Deepfaked audio often results in vocal tract reconstructions that resemble drinking straws rather than biological vocal tracts.

Deepfaked audio often results in vocal tract reconstructions that resemble drinking straws rather than biological vocal tracts.

Finally, Deepfakes are currently a concern, but over the next few years, they're going to get worse.

I hope in the long run, we can be able to combat this danger that is in our midst!









RELATED STORIES

L O A D I N G
. . . comments & more!
Hackernoon hq - po box 2206, edwards, colorado 81632, usa