A court document published by Forbes magazine showed how a complex bank robbery took place in Dubai, UAE, in early 2020.

The bank manager received a phone call from another manager known to him, and the manager claimed that the company was about to make a takeover and needed the bank to allow it to make some transfers of $35 million, and showed the bank manager emails from the lawyer and the person in charge of the deal.

All seemed to be fine;

So the bank manager completed the transfers, not knowing that he was part of a giant scam;

Its scammers used "deep voice" technology to reproduce the principal's speech.

This is the second time fraudsters have used such tools to commit fraud;

The first incident occurred in 2019, in which fraudsters impersonated the CEO of a UK-based energy company.

Jake Moore, a cybersecurity expert at information security company ESET, told Forbes, “Audio and visual deepfakes represent a fantastic development for 21st century technology, but they are also potentially incredibly dangerous, and pose a major threat to data, money and business. ".

"We are currently faced with active actors shifting expertise and resources to use the latest technology to manipulate people who are innocently aware of the potential or even existence of deepfake technology," Moore added.

And this technology is already advancing, becoming more and more prevalent;

In May 2021, technology company Veritone announced that it aimed to monetize deepfakes that would allow celebrities to present spoken programs without a word.

However, deepfakes come with a number of concerns and caveats.

For criminals, this is an easy process, all they need is access to audio samples (Getty Images)

How is fraud done?

For criminals, this technique is an easy process, all the criminal needs to do is access the audio samples, then combine the samples to create a fake voice, and choose the best opportunity to exploit it, by asking the victim to take a certain action using the fake voice.

And the latest advances in deep neural technologies are fueling the growth of VC technologies and paving the way for the production of high quality fake voices that are indistinguishable to human beings.

Two of the most important technologies in deep speech synthesis are WaveNet, a vocoder developed by DeepMind in 2016, and Tacotron, a text-to-text algorithm. Words created by Google (Google) in 2017.

The power of voice-transformation technologies also includes DNN, the mapping of linguistic features to be represented as sound features.

On the other hand, Wavenet converts audio features into high-quality audio files.

And finally, all these powerful audio simulators are readily available to the public.

As evidenced by this theft, deepfakes can cause many problems by the wrong people, yet technology or development cannot be stopped;

So the best we can do is move forward and work on solutions to prevent the wrong people from using these technologies.