A “senior official” spoke to the bank director and asked him to approve a $ 35 million expense. Now it turns out that behind the call was a voice synthesized using neural networks and a text-to-speech engine. Authorities are investigating
In recent years, we have seen countless videos of Deep-Pike, with the help of ML technologies and computer vision, and there is no doubt that the level in them is increasing. Now Deep-Pike also reaches the voice in question, with technologies that allow anyone who wants to create spoken sentences in the voice of any speaker, without him knowing anything about it of course. You are now experiencing this in the flesh of a senior banker, as tens of millions of dollars have been fraudulently withdrawn from your bank account.
This is not a senior in the company.
A group of tech-savvy scammers managed to fraudulently withdraw $ 35 million from a bank in the United Arab Emirates when they forged the voice of a senior official from a company whose account was with the bank. The bank manager received a call from someone whom he identified as a senior in the company, because he knew their voice through their joint work, so he was not suspicious when it came to a $ 35 million acquisition of the company.
After the conversation between the two, the “senior” informed the bank manager that a lawyer he had hired would be in charge of coordinating all the transfers and this happened. The bank manager received emails from the lawyer, and who he believed to be the senior, including the various transfer requests of the large capital intended for the benefit of the “purchase”, and the money disappeared from the company account.
In a document filed with US law enforcement by the Emirates Police, and released by Forbes It is written that the people behind the fraudulent spending of the bank’s funds have imitated the voice of the company’s top executive using Deep Voice technology, but it does not detail exactly how this was done. From a test we did, it appears that Deep Voice is the technology that uses neural networks to perform text-to-speech (or TTS for short) developed by researchers from the research division of Chinese Internet giant Baidu.
Deep-Weiss had three incarnations, with the latest version released by Baidu in 2018. It is true that there is a diversification of the models created by Baidu researchers on GitHub, including For the third and most updated version of Deep Voice That can be run in PyTorch from Facebook. According to Baidu, with the announcement of the latest version of the model, half an hour of speech allowed to synthesize text in 2,500 different voices. Baidu’s name was said to be not officially mentioned, neither in the report nor in the official document sent by law enforcement agencies to the United States.
The document, however, shows that 17 different people were involved in a scam that ended up being withdrawn from the company’s bank account – whose name was not disclosed – and sent to various bank accounts around the world. The document also shows that $ 400,000 of this amount went to bank accounts located in the United States.
No one involved in the case has given an official response to Forbes on the matter, either the US bank to which, according to the emirates, part of the stolen money was destined or the lawyer whose name was fraudulently linked and resides in the United States.
Tags for the article:
Introvert. Beer guru. Communicator. Travel fanatic. Web advocate. Certified alcohol geek. Tv buff. Subtly charming internet aficionado.