Controversy Can facial recognition appear on smart locks?

- Nov 27, 2018-

Controversy Can facial recognition appear on smart locks?

Nowadays, the face recognition technology scene has moved from Hannover, Germany to a wider field. At present, face payment, station entrance, and admission to candidates will be applied to face recognition technology, and the era of brushing has arrived.

Beginning with the integration of fingerprint recognition on Apple's iPhone 5s, biometric detection has become a real epidemic, and Apple's latest iPhoneX has cut off fingerprint recognition and used facial recognition as the only biometric method.

What followed was a series of follow-ups from other vendors. After all, Apple's integrated facial recognition on mobile phones is closer to popular recognition. The initial face recognition can be unlocked with photos, which is already a joke.

However, in today's fingerprint recognition, it is necessary for smart locks to follow up with mobile phone manufacturers, and to add face recognition to smart locks to completely replace fingerprint recognition? The author believes that it is necessary to consider this issue from three perspectives: 1. Rejection rate; 2. False rate; 3. Identification motivation

It should be noted that, up to now, Apple's "face ID" identification method is the most advanced, and is the most likely face recognition solution that is most widely adopted by future smart locks. Therefore, the following facial recognition is based on Apple's face. ID" example.

First, the rejection rate

The rejection rate and the falsehood rate are two important indicators in fingerprint recognition. The false rejection rate, which is FRR (False Rejection Rate), is simply the probability that the correct information (such as fingerprints, faces) will be rejected incorrectly.

Shortly after Apple iPhoneX was launched, many people complained about the frequency of unlocking passwords on mobile phones, which greatly exceeded the previous use of fingerprint-recognition mobile phones. Many people complained that they could not be unlocked by facial recognition when lying on the side of the bed. , you can only enter a password to unlock. This is the result of the high rate of facial recognition rejection.

In fingerprint recognition, this problem rarely occurs. This is the difference between the face information and the information characteristics of the fingerprint.

Fingerprint recognition is to press the finger on the recognition module, and the fingerprint image on the plane usually does not change greatly. The image features are stable and the information is dense, so that the accuracy of the recognition after encryption is relatively stable.

However, the characteristics of the face are different. Although the "face ID" has evolved from 2D image recognition to 3D depth of field recognition in acquiring information, a neural network chip has been added to the processor. However, the facial features of each person are mainly formed according to the bones, and will change in different situations.

For example, let a person run on a treadmill, and then use the video recorder to record the information in the lower part. In addition to the expression package that shakes and shakes, you will find that the facial feature information of almost every screenshot will be different. Two screenshots will even make you think of two people. Humans sometimes don't recognize facial information from a person in different situations, let alone a neural network chip that just started.

The situation mentioned above does not include makeup, wearing glasses, wearing a mask, dark circles and eye bags caused by poor sleep. As for getting fat and thin, Apple claims to be able to adapt gradually in the process of change, and other 3D face recognition technology seems to have not achieved this effect, but this technology can be followed up very quickly.

In addition to the rejection rate, the falsehood rate is also a relatively controversial part of facial recognition.

Second, the rate of recognition

The false rate, that is, the FAR (False Rejection Rate), is also called the “false positive rate”. It is to judge the wrong information as correct and finally pass the certification. This kind of indicator is difficult to find problems in fingerprint recognition, because it is difficult for you to find two people with similar fingerprints, and the characteristics of fingerprints are not regular. Therefore, in fingerprint identification, the high falsehood rate is most likely to be related to the encryption algorithm. relationship.

In facial recognition, this falsehood rate has a big problem. The most prominent problem is the problem of similar faces caused by double (multiple) births. Someone has done this experiment not long ago, and one of several pairs of twins entered facial information through the iPhoneX, and then the other of the twins tried to unlock.

From the actual test results, the falsehood rate of this situation is as high as nearly 50%. Even more news is that two people who are completely unrelated in the country can unlock the same iPhoneX just because they look similar. Although the feature number of the "face ID" on the iPhoneX is much more than that collected by humans when viewing the face, in the twins, especially the identical twins who are very close in appearance, they will still be defeated.

Many people (including the author's face blindness) often distinguish between two twins, and the neural network chip that wants to make the algorithm not mature enough to distinguish between twins or two people with similar looks needs to be very long. Time to evolve on the algorithm.

However, it is gratifying that as face recognition evolves from 2D image recognition to 3D depth of field recognition, unlocking the phone with only photos or videos will no longer exist. This is also the advantage brought by the advancement of recognition technology. The rate of recognition of facial recognition has increased several levels.

However, this kind of mistake in facial recognition is placed on the smart lock, and the problem caused by the bad relationship between the twins is excluded. Most twins will welcome their fellow brothers into the house. Therefore, in the face of mature 3D depth of field facial recognition, the false recognition rate of smart locks should not be a big problem. After all, the difficulty in admitting the false rate is not to be cracked, but to find the person who looks like.

The really big problem is what to say next: identify motivation.

Third, identify the motivation

As for the motivation to identify, the word is what the author thinks, because no suitable term can be found to describe it. This is the most worrying issue when I just released the iPhoneX. Especially on the smart lock, the recognition motivation of facial recognition is the most worrying to the author.

In the unlocking method mentioned above, fingerprints and faces are just a key, just a means of authentication. The motivation to trigger recognition is the most important part of biometric identification.

An example of fingerprint recognition with a smart lock. When you stand in front of the door and wait for the lock to be unlocked, press your finger on the fingerprint identification module and you can open the door after verification. And if you don't want to open the door, how long do you stand in front of the door will not open itself.

But the facial recognition is different. You just walked to the door and the door "嘭" opened. If it is the building's infrared automatic door will feel very natural, but this is its own security door, it is very convenient to watch, in fact, the crisis.

For example, if you meet someone who is not good with you at the door, or someone who sells some kind of sales, you are definitely not willing to let the door open in front of others. Face recognition will detect the face of your face, the door of the "beep" is open, or just unlocked without opening the door.

But at this time the door is already unlocked, and anyone can enter the room without your permission. The door lock for fingerprint recognition is still closed when you don't extend your hand for verification.

This is not only a problem of facial recognition, but also a problem that almost every type of non-contact biometric recognition encounters, but the iris threshold is too low to be compared with the iris recognition. This also makes facial recognition less secure than recognition fingerprints.

In addition, the fingerprint smart lock can also set a temporary fingerprint, or an emergency fingerprint, so that different permissions and functions can be realized, but the facial recognition is relatively simple in function, especially the emergency alarm fingerprint, which requires other keys. After all, each person has only one face, so only one function can be realized.

So what can we do? Since facial recognition is not safe enough on smart locks, what can we do? How to do it? wait.

Yes, we can only wait for a technical breakthrough in facial recognition. Face recognition at this time is not mature enough even if it is led by an apple.

First, the reality we need to face is that facial recognition is not widely available. Not only is it a smart lock, it is not completely popular even in the field of mobile phones. Face recognition on today's smart locks is mostly limited to 2D image recognition through a single camera, and this kind of face recognition that can be fooled with a photo is very unsafe and is not recommended.

Some smart locks also use infrared camera technology similar to iPhoneX, but the processing power of the chip is not at a level compared with Apple's A11 biochip. The main reason is that the technical solution is not mature enough to wait for technical breakthrough. .

Secondly, there won't be a smart lock manufacturer that is as radical as Apple, using face recognition as the only way to unlock it except passwords. However, due to the cost reduction of the fingerprint identification module, it is believed that fingerprint recognition can be used as an alternative unlocking method for smart locks. After all, there are only one face and ten fingers, and fingerprint recognition can realize multiple functions and functions. It is much larger than facial recognition.

Finally, I believe that when face recognition is widely used, facial recognition will develop into a very safe and complete biometric technology, both from the public's perception and technically, and the cost will drop dramatically. At that time, we might also explore the next more convenient and reliable identification technology.