The generation AI expert Xu Bing said in an interview with Lianhe Morning Post that the current deep -fake technology can make the fraud gang based on the need to produce false content."Differential technology is far behind generating technology, and the accuracy rate is not satisfactory. I am also more pessimistic about the criminal problem caused by generating technology."

In Malaysia's six states elections last year, Lin Huixian, a member of the Malaysian Democratic Front (MUDA), known as the most beautiful candidate.Ask her to ask her to pay back the money immediately.

28 -year -old Lin Huixian is an environmental protection element and an active Instagram user with more than 100,000 followers.However, frequent sharing of her life in social media brings her trouble, and she believes that there are scam gangs to stole her personal information and cheat.

Lin Huixian said that he has repeatedly complained to Instagram as a fake account number, but has not received much attention.Instagram's parent company Meta spokesman told Lianhe Zaobao that the platform has reviewed relevant accounts and took action, but "bad actors will continue to change the strategy to evade monitoring, which also means that no tool can become a pitch."(Provided by the respondent)

Lin Huixian told Lianhe Zaobao: "This situation has become more and more recent. I receive 10 to 20 threat text messages every day.

In order to confirm her innocence, she reported the case to the police on January 31. She reported the case in public media publicly, and also issued a document in the social media several times, but the situation did not improve.

On January 31st, Lin Huixian reported the case at the Baisaro Police Department of Kuala Lumpur, saying that personal information was severely stolen by criminals, and the victims came from Singapore, the United States, Canada, Turkey, and Kazakhstan.(Provided by the respondent)

According to the screenshot of the text message provided by Lin Huixian, the reporter contacted several victims who claimed to be scammed by Lin Huixian.

Turkish man Yusov (29 years old) said in an interview that in December 2022, a woman who claimed to be "Li Chu Xue" established a contact with him by mistake, and then launched a sweet offensive to him.He and the other party started online dating for a year.During the period, the other party repeatedly used "to save more money to get married". He lobbying him to conduct cryptocurrency transactions, and finally he was deceived $ 91,000 (about S $ 12,200).

Yosov firmly believes that "Li Chu Xue" is Lin Huixian."I have a video call with her four times, the screen is high -definition, the sound is clear, and there are countless photos and videos. Her alarm statement proves nothing."

Lin Huixian refuted that she had never contacted Yosov, and she did not deceive the other party's money.

After Yosov found Lin Huixian's contact information through Google, he contacted the other party many times and asked her to pay back the money.(Provided by the respondent)

The other two Turkish men Azdemir (45 years old) and Sercan (32) who did not want to disclose the full name told the United Morning Post that they were deceived by "Suna" and "Dora".A lot of money, they also believe that "Suna" and "Dora" are Lin Huixian.They contacted other victims, and there were at least 27 people.They have organized and reported to Turkey, and the case has now entered the investigation stage.

On December 6, 2023, the scammer who claimed to be "Dora" contacted Selkan by mistake, and then persuaded Cekan to invest in "Uncle's Internal Trading".He was eventually deceived $ 8,000.(Provided by the respondent)

Alphonse Nguyen, a Canadian man, realized that he might be recorded in the "pig kill" scam and recorded the conversation with the fraud.

Alfaus said in an interview that when the cryptocurrency trading platform refused to let him withdraw money for the third time, he had an unknown premonition, so he asked for three weeks to talk about the three weeks of the girl "Olivia Wong" video call."I was engaged in computer -related industries. It didn't take long for the video calls to be cheated.

57-year-old Armen, as a prove, requires a video call with a close friend "Olivia Wong" (left), and found that there are differences in the video and photos, and they are aware of being deceived.(Provided by the respondent)

Experts: Seeing that it is not necessarily a vigilance for the real people

Three Turkish respondents do not believe that artificial intelligence (AI) Deepfake technology can make high -definition videos.Ozdmir said: "In the two minutes of video calls with each other, I even saw the acne on her face."

The generative AI expert Xu Bing said in an interview that the current deep-fake technology can make the fraud gang make a real forgery video according to needs.(Video screenshot)

Xu Bing, chief executive and co -founder of Hippoml, an AI expert and AI company, said in an interview with Lianhe Zaobao that the current deep -fake technology can make fraud gangs based on the needs and make faces with acne, freckles and other faces.False content of the details.

At the request of the reporter, Xu Bing analyzed the video recorded by Alofi.He pointed out that the picture taken by common cameras is rectangular, but the video is a square -default square of the generated model. "This is obviously problematic."

He said: "Because the opponent rarely has physical movements, it is not ruled out that the initial face change technology has been used."

Zhou, Associate Professor, Associate Professor of the School of Network Space Security, China University of Science and TechnologyWen Bai analyzed Alofeus's video screenshots that the people in the video lacked details, the texture was too smooth, the light and shadow changed insufficient, and the face would shake obviously from time to time."This is a typical picture generated by the effects of deep -fake models."

Zhou Wenbai is an expert in the field of falsification and AI security. He suggested that people must always be vigilant to avoid being deceived by deep fake content.(Video screenshot)

Zhou Wenbai suggested that people ask for the object of video calls to turn around, cover the face, squeeze nose, etc. to try to distinguish the authenticity of the people in the video.

But he emphasized that the important thing is not to find clearly available identification methods, but to improve alertness."Without the habit of maintaining alertness, you will not think of using those identification methods."

Although experts suggest that people observe the physical laws of the real world, such as observing the outline of the character, the sense of facial coordination, the number of blinks, etc. to distinguish the authenticity, the American artificial intelligence company OpenAI launched in February this year this yearText generates video model SORA, which can more accurately generate realistic videos that conform to the physical principles of real world, and once again shake the reliability of identifying authenticity with the eyes.

The law can't keep up with AI deep pseudo

What worsen the situation is that the identification technology that should be renewed and updated with the generation technology has fallen into the bottleneck of development.

Jungpil Hahn, a professor of information system and analytical professor of the National University of Singapore's National University, said in an interview that existing technologies, such as watermark monitoring tools of images and text, have certain operating defects, and the accuracy rate is not high, soA large number of "false positives" often appear in the automatic monitoring system.

Professor Han Yan said that at any time, malicious people may have an advantage because researchers must create better deep-fake training data to develop new testing technologies.(Take from the website of the National University)

Han Yan said: "In this way, many platforms can only hire professional examiners to identify authenticity, but the cost of this method is very high, and online platforms such as social media may not be able to bear it."

Deep pseudo and identification are obviously a technical competition you chase me.

Xu Bing said: "The identification technology is far behind the generation technology, and the accuracy is not satisfactory. I am also more pessimistic about the crime caused by the generation technology."

Experts' worry and pessimism are self -coming.The British Internet Verification Company Sumsub's 2023 identity fraud report earlier this year pointed out that from 2022 to 2023, the deep fake incident in the Asia -Pacific region increased by 1530%.It can be seen that deep pseudo is becoming a powerful criminal tool for fraud gangs.

In February this year, AI Godfather Yoshua Bengio signed an open letter with more than 400 experts and industry executives, referring to deep pseudo -technology or endangering society, and called for strengthening supervision of deep pseudo technology.

Han Yan said that the current laws formulated by governments of various countries can not effectively resolve deep and fake crimes, and promote new legislation and take a long time, which is not conducive to instant punishment of cyber criminal acts.

He proposed two regulatory solutions: First, the platform is required to actively screen for bad content, blocking and deleting, but the operating cost of this solution is high and it is not easy to promote the implementation of the platform.This is a passive solution, which has a timeliness, because the harmful content will continue to spread before being withdrawn, and the damage will continue.

When the AI ​​deep pseudo -fraud subverts the "seeing is true", experts suggest that people can only keep vigilance at all times, so as not to trust, disclose, and transfer the certification password with relatives and friends"Positive body".