This article is from the public number:CV Intelligence (ID: CVAI2019), author : Zhang Xue; title map from: Visual China.

In the few seconds before the end of the game time, when the judges and the host almost announced the failure of the festival, the audience suddenly found a picture of Huang Jianxiang who had “passive hands and feet” on the screen.With clear Ivanka photos, at the last minute, the TSAIL team completed the task of making AI mistakes.

The same dramatic thing is that in the challenge of fingerprint recognition, the contestants completed the fingerprint production within 20 seconds of the countdown time. However, in the process of taking the fingerprint to verify, the fingerprints were accidentally landed. Exclaimed. Fortunately, the fingerprint was found and four smart devices were successfully opened.

This is a real picture of CV Intelligence witnessing at GeekPwn2019.

A most dramatic fingerprint challenge

In real life, you may inadvertently leave a lot of fingerprints. These fingerprints look very “light” and very insignificant, but the information is enough for the criminals to make your fingerprint.

“Because time is not enough, the countdown will be forcibly stopped when it reaches 20 seconds. The fingerprint has not been carved, only 2/3 of the fingerprint is used, and another 1/3 has not been finished.” Just finished the GeekPwn2019 fingerprint Chen Yu, a player in the Tencent Xuanwu Lab fingerprint recognition project, who unlocked the challenge, said.

In the course of the challenge, Chen Hao experienced two dramatic moments, one was to complete the satisfactory fingerprint production in the last second before the deadline, and the other was in the process of verifying the equipment from the console. The fingerprint accidentally fell to the ground.

But this is 2/3 and because of accidentally falling off the fingerprint on the ground, it helped Chen Hao successfully cracked five devices in the fingerprint identification competition, and completed the challenge. It is enough to see that relying on technical means to obtain ordinary users. Fingerprint is no longer a difficult thing in the tech world.

Unlike other competitions, the fingerprint recognition challenge is only attended by Chen Hao alone.

At the beginning of the game, most of the audience were either chatting or chatting at noon.

Why are there no opponents? Facing the mediaAsked, Chen Yu explained: “There are two kinds of pwn and exhibition games in the great game. I belong to the exhibition game.”

At the beginning of the competition, the judges said that in this fingerprint unlocking challenge, the core point is to make a fingerprint within 20 minutes and open all the devices.

There are six main players who need fingerprint unlocking, four mobile phones and two fingerprint punchers. These six devices use different technologies.

At the event site, the volunteers randomly found the fingerprint on the glass. Then Chen Hao first got the target fingerprint on the glass, and then printed the fingerprint film with the 3D printer according to the fingerprint sample, thus unlocking the fingerprint.

Unlike the way of fingerprinting in the imagination, Chen Hao first put the glass in the dark box to take pictures – looking for fingerprints, but in the course of the challenge, this simple fingerprint capture action took nearly 1 minute.

Next, he magnified the photos in the phone and processed the noise in the image twice so that the fingerprints could be more accurate and clear in 3D printing.

In three minutes, Chen Hao completed the “3D printing” of the first finger film, but after a few minutes, he seemed to be dissatisfied with the photos taken. Then began a new round of repetitive actions.

As the game progresses step by step, more and more viewers have already slowed down from gossip and sleep, and began to concentrate on the game.

In the remaining 8 minutes and 30 seconds, Chen Hao made some cleansing of the traces of the sculpture and cut out the engraved fingerprints. In fact, this engraving process still takes a long time, about 5 minutes to complete.

Because it takes too much time in the fingerprint engraving process, the time left for fingerprint reprocessing is very limited.

Next, the terrible scene also happened. Since Chen Hao had not obtained a satisfactory fingerprint engraving image in the first half of the game, he did not have a fingerprint that could be cracked 20 seconds before the end of the game.

Until the time is running out, his carefully crafted fingerprints are finally formed.

“You can see that Chen Hao’s hands are shaking now, and he is sweating.” Moderator Huang Jianxiang snarled at the scene.

Before unlocking, there is another detail that pushes the audience’s attention to the climax, that is, the fingerprint is off the ground.

When this scene happened, Huang Jianxiang repeatedly stressed, “I promise thisBr>

Because of the excitement at the scene, they were very calm and rational during the interview. It may be that they have become accustomed to the results of the championship, and they have not shown obvious joy.

Until about the technical principles of the game, they began to linger.

In fact, now foreign AI has been able to achieve “face change”. However, from the principle point of view, the previous “face change” is still different from this challenge.

In this regard, they explained: “We didn’t change our face in the whole photo. We just made a slight change in the original photo. People don’t seem to change much, but this tiny change can make it The machine mistakenly thought it was a photo of Ivanka.”

” From an academic point of view, it is actually an optimization problem in mathematics. Because there are some loopholes in the neural network, optimization is a way. Through some optimization methods, the model is considered to be the target object to attack. Wanka. From the content of the challenge, different genders, nationalities, races and ages will have a certain impact.” They further talked about.

Generally speaking, offensive and defensive games are mainly divided into black box and white box from the technical point of view. White box refers to the specific parameters given by the organizer. In this way, the break will become easier.

The black box does not give specific parameters. It needs to find specific parameters with similar anti-push technology. The face recognition of this attack is a typical black box scene, which is more difficult.

“This face recognition attack mainly attacks Amazon and Microsoft’s API. For us, because of business confidentiality, we don’t know what it is like, but we find that different neural networks have something in common. Point, by exploiting these commonalities, they have common vulnerabilities, so we use these common vulnerabilities to attack,” the contestant said.

Rome is not built in one day. The technical accumulation of the players and the success of the competition are not dependent on the day and night.

They admit that the laboratory itself has accumulated a lot in this field. Specific to them personally, at least two years ago began research, wrote a lot of code, read a lot of papers. This time, we have also explored some new methods for the new competitions. The most recent month is mainly for task debugging.

So it seems to be expected to get good grades, and the failure of the game is unexpected.

Innovating in accumulation, breaking through in innovation, this series of actions eventually made TSAIL team finally scored 229.77 points.>

When asked about the feeling of mistakes, he replied, “There are some accidents that can’t be avoided during the game. But you can’t be nervous, calm down.”

It is gratifying to note that although the competition is mostly personal, there are many assists between young geeks. In the face of competition, their mentality is more about learning from each other and finding ways to improve in the competition. Huang Sizhen even said informally, “Friendship first, competition second.”

It is reported that as early as GeekPwn2018, the youth group has emerged in this geek competition. In the Machine Agent Challenge of the year, the only high school student team successfully won the third place in the competition with excellent performance.

Compared with previous years, in addition to the lower age limit, there are still many teams led by the “Technology Girl”.

Although it is still young, it is undeniable that many 00 and 10 have already won awards in many robot competitions at home and abroad. The new generation of young geeks is on the rise.

Conclusion

At this year’s GeekPwn2019, CV can clearly feel that with the popularity of cloud computing, AI, Internet of Things, 5G and other cutting-edge technologies, the influence of geeks is widening.

No matter whether it is fingerprint recognition in the game, face recognition or stealth surgery in the AI ​​era, while bringing people the cutting-edge technology experience, it also buryes the dangerous seeds.

However, fortunately, there are such a group of people in real life. They are like firefighters in cutting-edge technology. They have been exploring and exploiting loopholes and exploiting loopholes.

As ordinary people, they are shining on the field, relying on their own efforts to guard the safety of one side.

At the same time, the rise of a new group of young geeks also allows us to see the future world, the security can look like.