What should I do after "seeing" is not necessarily "believing"?
Following the previous incident of "being deceived by AI face changing for 4.3 million yuan in 10 minutes" attracted attention, on May 25, another AI face change fraud incident that occurred in April was exposed by the media. This time the amount reached 2.45 million, and the whole scam process took even shorter, only 9 seconds.
The day before, the Internet Society of China issued a reminder that criminals use AI technology to fake the voice of a specific person through voice synthesis, change faces through AI, disguise themselves as a specific person, and make video calls with others in real time. The replaced facial expression is natural, fake and real, and can impersonate others to contact the victim, win the victim's trust and commit fraud. In the face of new scams using AI technology, the general public needs to be vigilant and strengthen prevention!
01 9 seconds were deceived by 2.45 million, "made a video call and are acquaintances"
According to Dawan News reports, in April this year, a fraud case of "impersonating an acquaintance" occurred in Anqing Economic Development Zone, Anhui Province, and an investigation by the police of the Anti-fraud Center of the Public Security Subbureau found that the fraudsters used a 9-second intelligent AI face-changing video to pretend to be "acquaintances" to make victims relax their vigilance and commit fraud.
"In the video, I saw his face", "but it hung up in a few seconds, saying that it was inconvenient to be in a meeting." On April 27, Mr. He's WeChat "friend" suddenly launched a video call to him, and after the call was connected, Mr. He saw that the "friend" was in a conference room, and just when he was about to ask further, the "friend" directly hung up the phone and said that there were important things to explain in the meeting and that Mr. He needed to add QQ communication.
Subsequently, the "friend" told Mr. He on QQ that there is currently a project bidding that requires working capital, and hopes that Mr. He will help advance the payment. "Because I made a video call and was an acquaintance", "I didn't think much about it, so I transferred the money", based on the trust of the "acquaintance", Mr. He did not hesitate, and immediately asked his family to transfer 2.45 million yuan to the account designated by the other party, until he called the other party afterwards to learn that he was deceived.
After discovering that he had been deceived, Mr. He quickly called the police, and after receiving the report, the Anqing Economic Development Public Security Sub-Bureau immediately dispatched capable police forces to form a special team to inquire and track the flow of funds and information in the accounts involved in the case as soon as possible. On April 28, the special police rushed to Beijing overnight to arrest three suspects involved in fraud. On May 22, the police promptly returned the 1.32 million yuan recovered to Mr. He, and the case is currently under further investigation.
02 How to prevent AI face changing and voice change fraud? Call back and observe the screen
On May 24, the Internet Society of China issued an announcement that in the face of new scams using AI technology, the general public needs to be vigilant and strengthen prevention. According to the Internet Society of China, with the open source of deep synthesis technology, deep synthesis products and services have gradually increased, and illegal acts of fraud and defamation using fake audio and video such as "AI for face" and "AI voice change" are common.
The Internet Society of China said that in the AI era, text, sound, images and videos may be deeply synthesized, and in typical scenarios such as transfer and remittance, capital exchanges, it is necessary to verify and confirm by calling back the other party's mobile phone number and other additional communication methods, and do not directly transfer and remit money without verification based on a single communication channel, no matter who the other party is.
According to interface news, in response to the use of AI face changing and onomatopoeia technology to commit fraud, Li Di, CEO of Xiaoice Company, said that the fraud team that has been learned to use AI face change is mainly using Wav2Lip, an open-source technology, to drive lip shapes through audio. "But there are problems with some lip shapes, such as the wrong mouth shape when pouting, the edges are shaky, or the mouth shape does not match the facial features, etc."
Xiao Zihao, co-founder of Relais Wisdom and algorithm scientist, also said that during a video call, you can ask the other party to do some actions such as shaking your head and opening your mouth, "If the model made by Black Production is not too fine and only makes a face, the teeth or tongue generated when suddenly showing teeth will not be particularly clear, and abnormal shadows may appear at the ears." ”
According to Xiao Zihao, from a technical point of view, there are also traces to identify whether the content has been deepfake, and the identification ideas can be roughly divided into two types, one is to find image editing traces, and the other is to judge whether the behavior is in line with common sense.
03 Artificial intelligence fraud has been repeated one after another, and higher-level supervision urgently needs to be implemented
According to the Securities Times, the "godfather of artificial intelligence" Jeffrey Hinton warned earlier this year that the threat posed by artificial intelligence to mankind will be more urgent than climate change. Hinton believes that artificial intelligence is an intelligent tool that can assist humans in doing tedious, repetitive and creative work, but there will also be criminals who use it to do illegal things. Artificial intelligence can not only write articles, but also generate images and videos. If it is used to deceive, it can almost be chaotic, and it is difficult for ordinary people to identify it. If someone uses it to make false images, videos, and misinformation, it is far more destructive than pure text form.
According to Xinhuanet, "in just 10 minutes, I was deceived of 4.3 million yuan" and "changed my face and pretended to be a red actress to bring goods, and I made a lot of money"... This kind of "face changing" and onomatopoeia are unpreventable. All this shows that the abuse of face swapping has become a threat to the public interest.
In response to AI fraud incidents with increasingly updated technical means, on November 25, 2022, the Cyberspace Administration of China, the Ministry of Industry and Information Technology, and the Ministry of Public Security jointly issued the Provisions on the Administration of Deep Synthesis of Internet Information Services (hereinafter referred to as the Provisions), which came into force on January 10, 2023. The provisions require deep synthesis service providers to add identifiers that do not affect the use of information content generated or edited using their services. Where services that generate or significantly change the functions of information content, such as intelligent dialogue, synthetic human voices, face generation, immersive immersive immersive scenes, etc., shall be conspicuously marked to avoid confusion or misidentification by the public. It is required that no organization or individual use technical means to delete, tamper with, or conceal relevant marks.
On April 11 this year, the Cyberspace Administration of China drafted the Measures for the Administration of Generative Artificial Intelligence Services (Draft for Comments), which states that AI content must be truthful and accurate, and also explicitly mentions that illegal acquisition, disclosure, and use of personal information, privacy, and trade secrets are prohibited.
To break the situation, on the one hand, we must strengthen public education and establish a new awareness of anti-fraud and anti-fraud; On the other hand, it is more important to accelerate the promotion and application of defensive and adversarial technical tools. For example, you may wish to try embedding voiceprints and face recognition identity verification modules in commonly used social media software.
Source: Daily Economic News