Deep Fake Prevention in the Era of Generative AI

Era of Generative AI

Introduction to Deep Fake Technology

Deep fake is a portmanteau word in which “deep” means deep learning and “fake” means fake. So, deep learning means the advanced level of artificial intelligence that uses multiple layers of machine learning algorithms for the extraction of higher-level features from raw input data. It is also capable of learning unstructured input data such as human faces etc.

In this article, you will learn how deep fake technology is impacting the world due to its use in creating fake images and videos.

How Deep Fake Technology Works?

AI is also able to collect data based on your physical movements. This data is further processed for the creation of deep fake videos and images with the help of a Generative Adversarial Network (GAN) which is another kind of special machine learning. 2 neural networks compete with each other in learning the features of the training set that can be photographs of faces. Then new data is generated with the same characteristics; new photographs.

As the neural networks keep testing the images, so the fake images become highly convincing and look real. This makes the deep fake a highly potent threat. Additionally, GANSs can also fake other kinds of data besides videos and images. In some cases, deep fake machine learning techniques can also be used to create fake voices.

Examples of Deep Fake Technology

Examples of high-profile deep fakes are not difficult to find. For example, a deep fake video made by actor Jordan Peele in which he used the real footage of Former U.S. President Barack Obama and merged it with its impression of Obama to issue a warning against deep fake videos. He also showed that he merged the 2 halves to make a new deep fake video.

            Another deep fake video of Facebook CEO Mark Zuckerberg appears to talk about how FB controls the future by stealing user data, especially on Instagram.

           However deep fakes with less fake effects also make remarkable impacts for example Nancy Pelosi’s drunk video scored millions of views on YouTube. It was a fake video that was made by slowing down the video artificially to create slurring sound effects.

Threats of Deep Fake –Blackmail and Frauds

 Mostly deep fake videos are used for political purposes and also for personal revenge. But now they are increasingly being used for blackmailing and frauds. The CEO of a British energy firm also became a victim of a deep voice fake and tricked out $243,000 of the head of the parent company by requesting an emergency money transfer. This fake was so real that he did not consider checking it and funds were transferred to a third-party bank account. CEO became suspicious when his boss requested again for another transfer. This time alarm bell rang but now it was too late to refund the amount that he had already transferred.

Another case of fraud happened in France where instead of using deep fake, the fraudster used impersonation with meticulous copies of the office of foreign Minister Jean Yves le Drian to defraud the senior executives for millions of Euros. Fraudster Gilbert is now alleged to disguising himself as a Minister to ask money from executives of a company and wealthy persons to liberate the French Hostage in Syria.

It is also possible that deep fake videos can blackmail the presidents of companies by threatening them to publish the damaged deep fake images and videos unless they pay them. There is also another way of fraud that intruders can trick the employees by asking for privileges and passwords as the Chief Information Officer and hackers can hack all the sensitive databases. 

Deep fake videos and images are already used to blackmail famous female journalists and reporters such as Rana Ayub in India who expose the abuses of power.

How You Can Protect Yourself Against Deep Fake?

To address the threats of deep fake, legislation has already begun. For example, the state of California passed 2 bills last year to make aspects of deep fake illegal; AB-730 banned the image manipulation of political candidates within sixty days of the election and AB-602 banned the usage of human image synthesis for pornography without the permission of depicted people.

Fortunately, cyber security companies are also trying to come up with better-detecting algorithms of Generative AI that help spot the tiny distortions and videos that are created in the faking process.

Features to Spot the Deep Fake Video

Deep fake videos are still at the stage where you can spot the signs of their fakeness by yourself. There are some characteristics of deep fake videos;

  • Shift in skin tone.
  • Image with digital artifacts.
  • Jerky Movements in the video.
  • Poor lip-synching with speech.
  • No blinking or strange blinking.

As deep fake technology improves, you will not be able to detect it with your eyes. You need to make a good cyber security program.

Effective Security Procedures for Protection against Deep Fake

save a life
  • Educate yourself and others about deep fake detection.
  • Make sure you use good quality media news sources.
  • Educate your family and employees about working on deep fake and its challenges.
  • Use the basic protocol “Trust but Verify”.
  • Regular data backups prevent tour data from running somewhere and provide you facility to restore data.
  • Use strong and different passwords for different accounts. For example, if someone gets into your Facebook account, you would not like to able them to access your other accounts.

Leave a Reply

Your email address will not be published. Required fields are marked *