Your face, 5 cents

"I suddenly feel so sad. It turns out that my handsome face is only worth 50 cents!"

Yesterday, a piece of news became a hot topic on Weibo. A Xinhua News Agency reporter found that there was a large amount of facial information being sold in packages on the online black market, and the price was only 0.5 yuan per copy.

The jaw-dropping thing is far from this. On a certain online trading platform, by searching for specific keywords, you can also find a store selling "photo activation" tools.

Someone asked: What is "photo activation"?

For example, when you buy a financial product, your phone will prompt you to perform a "face recognition" authentication operation. This operation is roughly divided into two steps:

1. You are required to point your mobile phone camera at yourself and take a static photo of the front for system review.

2. The system requires you to point the camera at yourself again and follow the instructions to blink, shake your head, open your mouth, etc. to complete the recognition.

The purpose of these two steps can be explained in plain words as follows: the machine is judging that you are you.

Here comes the point. By using the "Photo Activation" tool, others can use your static photos to perform dynamic actions such as "blinking, opening mouth, nodding", etc.

Interestingly, if the buyer does not know how to operate this "photo activation" tool, the seller also provides a teaching service that provides the software and tutorial for 35 yuan, and the buyer will receive a teaching link after payment.

An IEEE Fellow mentioned in an interview with AI Gold Rush that the implementation of this type of "technology" is actually not difficult. 3D printed molds can deceive most three-dimensional face recognition systems; videos generated by GAN can deceive most two-dimensional face recognition systems.

Some people might be terrified when reading this, and they might even break out in a cold sweat. Others might ask: Even if we buy this facial information, what can we do?

This issue needs to be discussed in two situations:

1. If only your facial information is leaked, this is not a big problem;

2. If, in addition to your facial information, sensitive data including your ID number, bank card number, mobile phone number, etc. is also leaked, you need to be vigilant at all times.

With this fake facial information, criminals can exploit system loopholes to deceive the facial recognition machines of some platforms and then carry out illegal activities such as false registration and telecommunications network fraud.

In an interview with Xinhua News Agency, a seller even boasted, " As long as you are proficient in using the toolbox, you can also use this facial data to help unblock frozen WeChat and Alipay accounts. "

This means that when you are living a peaceful life, the money in your WeChat and Yu'ebao may have been transferred away; when you are having fun outside, your home's fingerprint lock may have been opened and the treasure you placed in the express cabinet may have been taken away.

In addition to the toolbox, at the end of last year, according to a report by Fortune magazine, Kneron, an artificial intelligence company in San Diego, USA, used a special 3D mask to successfully deceive several famous facial recognition payment systems in the world and completed the shopping payment process.

The team also claims that they even entered train stations in China using the same method.

Kneron CEO Liu Juncheng said this shows that facial recognition technology does not meet security standards, which will pose a threat to user privacy.

"If facial recognition is so unsafe, let's use fingerprint recognition instead!"

Sorry to tell you, there may be problems.

Last September, an expert from an industry association stated that if the lens is close enough when taking a photo, the "scissor hands" photo can restore the fingerprint information of the person in the photo through photo magnification technology and artificial intelligence enhancement technology.

"Basically, photos of the hand gesture taken within 1.5 meters can restore 100% of the fingerprints, photos taken within 1.5 meters to 3 meters can restore 50% of the fingerprints, and only photos taken over 3 meters away are difficult to extract fingerprints."

After the fingerprint is extracted, it is made into a fingerprint film using professional materials, which can be used by criminals in various channels that use fingerprint technology to identify identity, such as fingerprint door locks, fingerprint payments, etc.

Once this statement was put forward, it continued to ferment and was listed on the Weibo hot search list many times.

"Are we going to be invisible forever?"

Having read this far, many of my friends will probably sigh to the sky.

In fact, there is no need to worry too much. If you want to crack information at this level of financial security, not only will you run the risk of being arrested, but the procedure is also extremely complicated.

For example, before making the 3D mask mentioned above:

• First, you need to use a high-precision 3D scanner to scan the face and obtain three-dimensional data;

• Secondly, a high-precision 3D printer is needed to print out the main structure of the human head;

• Most importantly, millimeter-level restoration of facial features is required, including high-level restoration of skin color.

On the device side, if a user has activated face payment, if he wants to use other mobile phones for facial recognition payment or transfer, he must re-activate face payment using the payment password.

In addition, when logging in using other mobile phones, you also need to perform login identity authentication, which means you need a login password.

In other words, if you want to use a 3D mask to steal credit cards, you must meet several conditions at the same time:

1. Collect high-precision 3D models of Alipay users’ heads;

2. Complete the production of masks with millimeter-level precision regardless of cost, and strive for perfection;

3. Get the user’s mobile phone if the Alipay user has activated face-scanning payment, or directly obtain the Alipay user’s login password and payment password.

Obviously, in actual operation, this method of stealing credit cards is not feasible.

Regarding the issue of fingerprint recognition security, industry expert Jiu Di Shui said, "There is no need to worry at all."

He mentioned that fingerprints are the raised lines on the epidermis. Since human fingerprints are the result of the combined effects of genetics and the environment, everyone has fingerprints, but they are different.

Since the repetition rate of fingerprints is extremely small, about one in 15 billion, it is called a human ID card.

If you stretch out your hand and observe carefully, you can find that tiny fingerprints can also be divided into three types: some concentric circles or spiral lines look like whirlpools in water, which are called bucket-shaped lines; some lines are like winnowing baskets with one side open, which are called dustpan-shaped lines; some lines are shaped like a bow, which are called bow-shaped lines.

In addition to the different overall shapes, the number and length of each person's fingerprint patterns are also different.

If you enlarge the fingerprint, you can find different detailed features on the fingerprint, which are divided into 9 types: divergence; combination; small eye; small dot; end point; short stick; hook; small bridge; starting point.

At present, all scientific and technological achievements based on fingerprint recognition, whether it is fingerprint matching at the crime scene, Alipay fingerprint payment, or mobile phone fingerprint unlocking, all utilize these 9 detailed features on the fingerprint.

Generally, after humans reach adulthood, the structure and shape of fingerprints remain basically unchanged, and the detailed features of the fingerprints on each finger remain unchanged.

The size, shape, and two-dimensional coordinates of detailed features are relatively fixed, which is also the premise for people to do fingerprint matching. If the detailed features of a fingerprint cannot be seen clearly, then this fingerprint will be called a "waste fingerprint" and cannot be used in any field.

There are many ways to extract fingerprints at a crime scene, the most common ones include powder brushing, reagent development, multi-band light source and other methods.

After these fingerprints appear, they need to be photographed and fixed. The equipment used for taking photos is mostly high-resolution fine-eye lenses, which are expensive. In addition to the lenses, special light sources must be used to take clear enough photos.

At the same time, when taking fingerprints, a scale must be placed, and after multiple focusing, ultra-close-up shooting is required to obtain a clear fingerprint image sample for comparison.

Finally, during the comparison process, the fingerprint must be magnified large enough and clear enough to observe the detailed features, which is also the key to fingerprint identification.

In summary, four conclusions can be drawn:

1. When you use a mobile phone to take a picture of your fingerprint, even at a very close distance, you cannot capture the detailed features of the fingerprint. The principle of current fingerprint locks and fingerprint recognition devices is to recognize the 9 detailed features on the fingerprint. If the detailed features are not clearly captured, it will not cause any threat.

2. Detailed features need to be captured with a special lens, but this lens must be close to the object being photographed, and if necessary, it needs to be focused multiple times with the help of a special light source, otherwise the picture will not be clear at all; and it is useless to obtain only one or two detailed features on the finger. The fingerprint recognition on the mobile phone needs to record the detailed features of most fingerprints before it can be used.

3. Fingerprint photos without scale are useless. If the detailed features of the fingerprint are placed on a plane, not only its individual shape needs to be recorded, but also its two-dimensional coordinates at a fixed position (because the fingerprints of adults basically remain unchanged, so the detailed features are also relatively fixed).

For example, the nth line in the fingerprint has a "small eye" feature, and its distance to the XX line is how far. As long as the data is biased, it cannot be identified.

4. Fingerprint matching is a relatively complicated process that requires the use of special instruments and the participation of professionals.

The frequent occurrence of such information security incidents reminds people to do two things:

1. Always pay attention to the security of your personal privacy data;

2. Large technology companies need to be more responsible.

At the same time, for AI technologies including facial recognition, we should perhaps also give them some opportunities for trial and error and correction.

The wheel of time rolls forward, science and technology continue to advance, and everywhere we look are the results of human technological excellence.

People are always afraid of unknown things, but they cannot escape their temptation because they are greedy. Because of greed, human beings are always exploring and developing.

Everything exists dialectically, and rapid development will bring unexpected changes. But there is nothing wrong with technology itself, and we cannot throw the baby out with the bathwater or deny everything because of individual events.

Just like facial recognition, the implementation of this technology has brought people infinite convenience and excellent experience.

After all, there is no absolute security, and Rome was not built in a day.