Shulou(Shulou.com)11/24 Report--

AI face change, once only bothered celebrities, but now it will directly affect the property security of our ordinary people.

There is not a small number of fake face theft accounts and money transfers. In August this year, two suspects were arrested by the police for using mobile phone software to make dynamic face recognition videos and "disguise" to log on to other people's online accounts to steal funds.

According to news reports and court rulings, the illegal profits of such cases range from thousands to hundreds of thousands of yuan, and many similar cases have taken place all over the country. Some lawbreakers even use the bank's face recognition authorization function to steal deposits by intercepting SMS verification codes through Trojans, with a total amount of more than 2 million yuan involved.

If you can't fool people, you can fool the camera! AI face changing, the term Deepfake, is a technology that combines Deep learning (deep learning) and Fake (forgery) and uses AI to change one person's face into another person's photo / video.

There are quite a number of "criminal applications" around AI face changing, including but not limited to:

Attack facial scan verification: cash out other people's information directly from Alipay, Wechat wallet, and even loan software.

Make fake pornographic pictures / videos: fraud, extortion or damage to the reputation of others

Real-time face-changing calls: stealing and fraudulent relatives and friends

Create false information: deceive politicians, judges, investors, etc.

Release fake news: incite people, cause chaos and panic, crack down on business rivals, create stock market turmoil, and so on.

Among them, facial scanning verification has the most important relationship with ordinary people, because it verifies a person's identity. Once this information is broken, your asset security and information security will be threatened, and it is easy to "streak" on the Internet.

In addition to Deepfake technology, there are quite a number of ways to interfere with face recognition authentication, such as creating a "cheat AI glasses":

I'm sorry I misplaced it. It's this:

Source: Accessorize to a crime: Real and stealthy attacks onstate-of-the-art face recognition. In a classic literature in the field of image recognition, the researchers designed a specific "glasses" (middle image) through mathematical operations, which can be worn to make one person (left) be recognized as another person by AI face recognition (right).

Another way to cheat detection devices is 3D masks, such as the following images:

Photo Source: businessinsider & kneron, a product of the company Kneron, announced at the end of 2019 that facial scan payments from Wechat, Alipay and face recognition gates at railway stations can all pass by disguised as someone else.

AI changed his face, where in particular? For ordinary people, the above technologies do not need to worry too much, because they have a variety of defects, such as "cheating AI glasses" is mainly aimed at static picture recognition, can not break through dynamic face recognition, while "face mask" needs to be customized, expensive and complex, so it is not cost-effective to use on ordinary people.

But AI has changed everything, reducing the cost of counterfeiting to the "cabbage price".

The shorter the video, the lower the pixels, the lower the real-time requirements (whether you need to change your face in real time or you can send it after making the video), the lower the vigilance of the deceiving object, and the more available material (that is, multi-angle photos and videos of the target). The lower the cost of counterfeiting.

For example, the same person talking, AI into another person's face, can be directly used to use the AI model with NVIDIA GTX1080 Ti can do this, the action is smooth, but the "plastic" is serious, you can see at a glance that it is not a real video.

On the other hand, the AI model trained by NVIDIA RTX 3090 Ti for a long time (hours or even days) can produce quite a real effect. Perhaps only after watching carefully and repeatedly, will you find out: eh? Is the wrinkle on this man's forehead flashing just now?

For "face recognition", deepfake is much cheaper; because to break "face recognition" does not require the most "fine", the most time-consuming Deepfake technology, only a few seconds of video with average definition.

On the other hand, more elaborate videos produced after increasing the cost can be used for other fraudulent purposes. And this kind of counterfeiting is a kind of "one-cost-for-ten-thousand-profit" business. After the equipment and algorithms are complete, fake videos can be mass produced according to different situations and needs.

What can ordinary people do when AI changes its face and enters the lives of ordinary people, what can we do to deal with the illegal acts associated with it?

What should I do to change the face of the machine? For scenarios such as facial scanning to verify identity, forged videos need to be automatically verified by the face recognition system, or "fooled" the machine. The video time required is short (within a few seconds), the dynamic is simple (only a small number of fixed actions, and even some systems only recognize pictures), and there are no real-time requirements, which are relatively simple in technology, and the recognition end is not under the control of a real person.

For such cases, what ordinary people can do is to protect their personal information (including facial information), and try to use a variety of ways of authentication.

Password, fingerprint, mobile phone number, optional verification methods are chosen, although they will also be in trouble, but equivalent to several lines of insurance, the possibility of all being breached at the same time is still much smaller.

If it is unfortunately stolen, call the police immediately, cooperate with the police to track down lawbreakers and recover illegal gains, and report to the platform at the same time, so that the platform can get more information to repair loopholes and upgrade the system.

"deceive" a real person to change his face, what to do? AI face-changing deceiving real people, such as in the face of personal extortion, fraud, and false news from the public, appears to be more hidden and diversified.

Fundamentally speaking, with AI face-changing technology, ordinary people should be more mindful in the face of image information, and cannot blindly believe that seeing is believing.

If you are blackmailed with fake nude photos or indecent videos, don't respond, just call the police.

The AI face-changing target of this kind of crime is the victim, so as long as the person who sees the video knows that he has not taken such photos or videos, he can judge that the video is forged. On the contrary, the biggest difficulty is not to be confused and not to pay money to lawbreakers out of fear.

If someone uses fake video fraud, such as a swindler to steal your friend's number, use AI to change your face and borrow money from you on the phone in real time. At this time, multi-channel verification is very important, various social platforms, mailboxes, text messages, phone calls, do not want to transfer money just because the other person looks anxious.

When watching news and videos on the Internet, check whether the news sources are reliable. Especially what Biden speech, Musk speech, these celebrities can be used for machine learning training set is too many, it is easy to copy high-precision video.

Musk in this picture is a fake. | Image source: some tips for VR gyro to identify Deepfake and some tricks to help you better determine whether a video is forged by AI face changing (these methods are not very reliable if you open a 10-level beauty screen. After all, level 10 beauty can also be seen as an AI face change):

Pay more attention to the shape of the face to see whether the size, shape, jaw line position, especially the moving appearance of the face is consistent with myself.

Pay attention to wrinkles after all, everyone's skin condition and wrinkle direction are different, AI model only based on a few photos (and not necessarily recent photos), it is difficult to produce exactly the same wrinkles as myself.

A person has too smooth skin, too many wrinkles, or inconsistent skin conditions on the whole face (such as a lot of wrinkles on the forehead, smooth cheeks), and a different sense of age in a video (looking young and old). May be the characteristics of a fake video.

Pay attention to the background to see if this background is the background that this person usually uses, whether the connection between the background and people is natural, whether the background itself is out of shape, and so on.

The video generated by AI changing face does not follow the physical rules of light and shadow in the real world, so facial shadow, reflection of glasses, and so on, may sell fake videos.

Pay attention to the location and size of facial features AI forged video may appear facial features from big to small, position drift phenomenon.

Focus on unique facial features if the person has moles, tattoos, and scars, do they all look in the right place?

Does it look real to focus on hair volume and hairline? Does the edge of the hair look natural?

Paying attention to movements such as whether the frequency and mode of blinking is normal, whether the movement of eyebrows and mouth is the way this person normally looks, whether there is deformation when turning the head (especially to the 90-degree side), and whether the occlusion is clearly visible when the face is obscured, these are some of the key points to identify AI forged videos.

Some easy-to-reveal actions | Image Source: metaphysic & VFXChris Ume in addition to the above theoretical knowledge, you can also try to do practical exercises on this website (https://detectfakes.media.mit.edu/) to improve your ability to identify fake videos, and may also help scientists resist the illegal use of AI to change faces.

As long as there are high potential illegal gains, lawbreakers will not hesitate to invest more to make more beautiful and "real" fake videos to deceive ordinary people, big companies, large institutions and even the government.

Therefore, in addition to what individuals can do, it should also be continuously improved and strengthened by the government and companies in terms of policies, regulations and technological construction, so that technological progress can really benefit mankind, rather than becoming a criminal tool for harming others.

references

[1] http://epaper.ynet.com/html/2022-04/14/content_396257.htm?div=-1

Startled Tangmu | seeing is believing? "changing face" steals Alipay! Someone has been hit. (2022). Retrieved 25 October 2022, from https://mp.weixin.qq.com/ s / qNEaUN-A_83OL53vHvbWww

[2] https://www.theverge.com/2019/12/13/21020575/china-facial-recognition-terminals-fooled-3d-mask-kneron-research-fallibility

[3] Kneron's Edge AI & Facial Recognition Survey Pushes Forward Industry | Kneron-Full Stack Edge AI. (2020). Retrieved 25 October 2022, from https://www.kneron.com/news/blog/85/

[4] All it takes to fool facial recognition at airports and border crossings is a printed mask, researchers found. (2022). Retrieved 25 October 2022, from https://www.businessinsider.com/facial-recognition-fooled-with-mask-kneron-tests-2019-12

To Uncover a Deepfake Video Call, Ask the Caller to Turn Sideways-Metaphysic.ai. (2022). Retrieved 25 October 2022, from https://metaphysic.ai/to-uncover-a-deepfake-video-call-ask-the-caller-to-turn-sideways/

[6] Detecting deepfakes by looking closely reveals a way to protect against them. (2022). Retrieved 25 October 2022, from https://phys.org/news/2019-06-deepfakes-reveals.html

[7] Europol report finds deepfake technology could become staple tool for organised crime | Europol. (2022). Retrieved 25 October 2022, from https://www.europol.europa.eu/media-press/newsroom/news/europol-report-finds-deepfake-technology-could-become-staple-tool-for-organised-crime

[8] Project Overview < Detect DeepFakes: How to counteract misinformation created by AI-MIT Media Lab. (2022). Retrieved 25 October 2022, from https://www.media.mit.edu/projects/detect-fakes/overview/

[9] Sharif, M., Bhagavatula, S.S., Bauer, L.A., & Reiter, M. K. (2016, October). Accessorize to a crime: Real and stealthy attacks on state-of-the-art face recognition. In Proceedings of the 2016 acm sigsac conference on computer and communications security (pp. 1528-1540).

[10] Thies, J., Zollh ö fer, M., Theobalt, C., Stamminger, M., & Nie é ner, M. (2018). Headon: Real-time reenactment of human portrait videos. ACM Transactions on Graphics (TOG), 37 (4), 1-13.

[11] Li, Y., Yang, X., Wu, B., & Lyu, S. (2019). Hiding faces in plain sight: Disrupting AI face synthesis with adversarial perturbations. ArXiv preprint arXiv:1906.09288.

[12] Westerlund, M. (2019). The emergence of deepfake technology: A review. Technology Innovation Management Review, 9 (11).

[13] Mirsky, Y., & Lee, W. (2021). The creation and detection of deepfakes: A survey. ACM Computing Surveys (CSUR), 54 (1), 1-41.

[14] Li, C., Wang, L.A., Ji, S.A., Zhang, X., Xi, Z., Guo, S.S. & Wang, T. (2022). Seeing is living? Rethinking the security of facial liveness verification in the deepfake era. CoRR abs/2202.10673.

[15] Nguyen, T. T., Nguyen, Q. V. H., Nguyen, D. T., Nguyen, D. T., Huynh-The, T., Nahavandi, S.,. & Nguyen, C. M. (2022). Deep learning for deepfakes creation and detection: A survey. Computer Vision and Image Understanding, 223, 103525.

[16] be vigilant! Forged "face" to steal the account, two men in Zhangzhou were arrested _ Tencent News. (2022). Retrieved 25 October 2022, from https://new.qq.com/ rain / a/20220830A068U500

This article comes from the official account of Wechat: fruit Shell (ID:Guokr42), author: Luowan, Editor: Emeria, Youyou

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.