According to the Wall Street Journal reported that a British energy company suffered a "AI language fraud," criminals using AI language generation software, posing as CEO of its parent company, British Energy CEO to call, require the energy company to within 1 hour another suppliers play money.
The British energy company's CEO, said in a telephone each other's accent, rhythm and parent company CEO exactly the same, so the victim immediately paid about 24 million dollars (about 1.73 million yuan) in accordance with the requirements of criminals, victims had no skeptical, but after a short time after the criminals receive the money once again called me and requested money again as soon as possible.
The victim feels a bit wrong, and finds himself suffered a fraud, and money has been fully transferred to Mexico, after verification. As of today, police have not yet announced whether to criminals, the case has not released more information, such as specific companies suffered fraud.
But for the deception of British Energy, the midst of sadness is that the insurance companies are willing to lose the full amount of the money.
Fraud tools upgrade
Pindrop (Security Technologies, Inc.) in 18 years, the report said that in 2013--2017 years, the voice of fraud class of cases increased by 350%, and in each of the 638 through voice frauds have one is the use of fraud class AI software to dial .
And in early July, Israel National Network Agency (INCD) issued a "new network attack" warning that criminals are likely to make use of AI techniques to imitate management, funds transfer and other command staff.
With this successful fraud criminals, I am afraid there will be a lot of copycat crime in the short term.
In case of AI technology to imitate the sound of criminal gangs based, can make a fool-operation, based on a very small number of specific samples, "a key change sound" of the software, I am afraid that time will suffer more than the management, we are also likely these people become victims.
We dare to even imagine, if the criminal gangs dedicated to making a small red envelope collar language program, designated induce you to say words like "I have an urgent need, give me a transfer / remittance / X dollars to fight money", this training model probably lead to "one key change sound" is not far away.
Blame open source?
There overseas users believe that the criminal use of AI techniques for fraud, largely because other open source on GitHub community there has been a large number of similar open source projects.
For example, mimic the actress Kate Winslet (Titanic care of the household) talking GitHub project "Neural_Network_Voices", a video presentation of the project is quite popular in the pipeline, but after all, imitation generation goddess Ruth. And got 6600+ star project "Real-Time-Voice-Cloning"
Prior to include millions of people reprimanded "a key strip" APP, technology behind the public are also a well-known open source projects.
Today more and more on technology tall, to be applied to every aspect of our lives, there are many which sparked huge controversy, such as the recent ZAO APP, supervised student teaching the class system, as well as public opinion behind the risk, so many people are in thinking about a problem: "how to prevent technology being spent to do evil?"
Can prevent, deter crime, only more advanced technology, more convenient use of technology, such as reptiles and anti-climb on, and derived a wide variety of code ......
Will continue the rapid development of technology will eventually, inevitably change the future, it might lead to greater conflict, but eventually will further improve our lives.
For those of us who are young, whether or not already entered the field of artificial intelligence, and now we are also personally create the future, change the future opportunities.
Now the most important thing is to seize the opportunity to catch the express train artificial intelligence technology into the realm of the most promising!
Sources and references:
https://www.reddit.com/r/MachineLearning/comments/cxskj1/d_fraudsters_used_ai_to_mimic_ceos_voice_in/
https://thenextweb.com/security/2019/09/02/fraudsters-deepfake-ceos-voice-to-trick-manager-into-transferring-243000/
https://www.cshub.com/attacks/articles/ai-could-escalate-new-type-of-voice-phishing-cyber-attacks