AI face-swapping fraud defrauded 4.3 million yuan. Immediately consumer experts recommend three defensive measures

Recently, the Baotou City Public Security Bureau announced that Mr. Guo, the legal representative of a technology company in Fuzhou City, was cheated out of 4.3 million yuan in 10 minutes due to AI face-changing, triggering widespread heated discussions on the safety of artificial intelligence.

In this regard, Feng Yue, an expert on intelligent core from the Masha Consumer Artificial Intelligence Research Institute, said, “The core functions of technologies such as in-vivo anti-counterfeiting and intermediary recognition are to

avoid the shortcomings of face recognition and effectively identify people. Photos, videos, masks, simulation models, etc., especially the face-changing and voice-changing fraud that occurred in this incident."

Regarding the potential risks brought by AI, "AI Godfather" Hinton once said in an interview with the New York Times, "Compared to climate change,

AI may pose a 'more urgent' threat to humanity.”

Based on reality, this statement is not “alarmist”. The "evil" in Pandora's Box is gradually emerging. At the same time, AIGC is rapidly promoting the development of AI to a new stage.

Beware of the “Evils” of Technology: Seeing Is Not necessarily Believing

According to the "Safe Baotou" WeChat public account, the Telecommunications Cybercrime Investigation Bureau of Baotou City Public Security Bureau announced a case of telecommunications fraud. Mr. Guo, the legal representative of a technology company in Fuzhou

was defrauded of 4.3 million yuan in 10 minutes. Yuan, fortunately, the 3.3684 million yuan of defrauded funds has been intercepted, and the "intelligent AI face-changing and onomatopoeia technology" was used. How did they do it?

This is a complicated process. From the news itself, from the perspective of Mr. Guo, the victim

was mainly deceived by face-changing technology and onomatopoeia technology, but there may be some other hidden things behind the incident. Conditions, such as through which channel the criminal established contact with the victim, how to obtain the account of the relevant social software, how to know who he needs to impersonate, how to obtain the corresponding image and sound materials, and how to select the victim.

If we focus on the two key issues of

face-changing technology and onomatopoeia technology, the main problems are obtaining basic data, training framework, and real-time rendering. At the data level, the audio and video information of the impersonator can be obtained from social engineering libraries such as social media, or it can be extracted from the photo album and chat history of a lost black phone. In terms of training framework, the editing software of several mainstream domestic short video platforms currently supports face-changing special effects without warning messages. There are more than 30 synthesis methods in foreign open source frameworks. In terms of real-time rendering, hijacked rendering can also be achieved through attack operations such as flashing the phone and installing intrusive software.

Therefore,

"Seeing is not necessarily believing". There is always deliberate "evil" behind technology, which must be guarded against.

Nip in the bud and defeat "magic" with "魔"

Risks are inevitable, so how can we use "magic" to defeat "magic"?

At present, many financial institutions have relevant implementation measures and related technology applications. For example, technologies such as

live anti-counterfeiting and intermediary identification have also been applied in financial business scenarios.

“The core function of live anti-counterfeiting, intermediary identification and other technologies is

to avoid the shortcomings of face recognition and effectively identify photos, videos, masks, simulation models, etc., especially those that appeared in this incident. Face-changing and voice-changing fraud." Feng Yue, an intelligent core expert at MaMaConsumer Artificial Intelligence Research Institute, said, "Based on such technological innovations, MaMaConsumer Finance is currently able to intercept 99% of batch fraud attacks and effectively prevent forged information. , intermediary agency, counterfeit applications, long-term lending, telecommunications fraud and other risks. "

Compared with financial institutions at the end of the overall process, social software is obliged to introduce a counterfeiting detection mechanism, and provide real-time call authentication reminders to users to mark whether the peer user has abnormal behavior or abnormal image forgery. , abnormal sound forgery. Earlier risk warning information can more effectively prevent the occurrence of risk results.

In addition to platform-related implementation, the public can also take some self-help measures. Currently, there is a type of technology that can protect user data from being used to change faces and voices. This type of technology is called

Adversarial Sample Technology. Taking human face images as an example, face-changing technology can be rendered ineffective by mixing an adversarial perturbation mask invisible to the naked eye into the image. Users can use this type of technology to protect their media information such as images and voices on social media before making it public.

Finally,

"precautionary awareness" is the cornerstone of all "magic", and the public also needs to strengthen the security protection of personal belongings and information.

"Don't click on unknown text message links, email links, etc., don't scan QR codes from unknown sources, download APPs, don't easily provide your face and other personal biometric information to others, don't easily reveal your ID card, bank card , verification codes and other information, and do not overly disclose or share animations, videos, etc.," Feng Yue suggested. "There is no guarantee that mobile phones and other mobile devices will not be lost. You can regularly clear WeChat chat records, hide relevant traces with people you have close relationships with, encrypt mobile photo albums, or activate the data protection function of your mobile phone. These daily behaviors may lead to fraud. Intercepted outside."

It takes a long time to achieve success, step by step. Preventing the "misuse" of artificial intelligence technology will remain a core issue in the digital era.

Source: Financial Industry Information

The above is the detailed content of AI face-swapping fraud defrauded 4.3 million yuan. Immediately consumer experts recommend three defensive measures. For more information, please follow other related articles on the PHP Chinese website!