The author is also very curious, is this cool "black technology" really as easy to use and safe as advertised? In the attitude that hearing is false and seeing is believing, we bought a new Xiaomi Note3 machine to understand this new function.
The first round: Is it easy to unlock the face?
Let's see if face unlocking is easy to use in various environments.
As we all know, illumination is one of the important factors affecting camera shooting and recognition, and it will inevitably become an important factor affecting the effect of face recognition. We tried the use of face unlocking in backlight and dark environment respectively.
Figure 1 Backlight Scene Figure 2 Dark light environment
Conclusion: After testing, Note3 can still unlock the face in backlight, but Xiaomi Note3 can't unlock the face in dark scene.
Then let's see if the face will fail to unlock if it is not completely exposed. We use sunglasses, masks, scarves and other items to cover some faces for testing.
Figure 3 Wearing sunglasses
Figure 4 Wearing a mask
Figure 5 Wearing a scarf
Conclusion: Xiaomi Note3 can't unlock the face without completely exposing it.
The second round: Is facial unlocking safe?
As early as this year's 3. 15 party, the host passed the verification of mobile phone face recognition by simulating faces with photos, which caused doubts about face recognition from all sides. So will Xiaomi Note3' s face recognition unlock have similar security vulnerabilities?
Figure 63. 15 parties
In order to verify whether Xiaomi Note3' s living defense technology is effective, we used color printed face photos, mobile phone electronic photos, iPad electronic photos and iPad videos for experiments, and found that the face photos of the above media really could not be unlocked by mobile phone faces.
However, the following further experiments surprised the author. The face recognition of Xiaomi Note3 is worthy of scrutiny!
First of all, we record the face of the tested person into Xiaomi Note3 through the face unlocking function, and verify that I can quickly unlock Xiaomi Note3 with human face.
Then we took a color photo of the tested person's face and printed it out, cut off more than a part along the edge of the head of the face, bent the cut face photo at a small angle and pointed it at the mobile phone, and the mobile phone was unlocked! ?
Fig. 7 cropped face photo
According to the above ideas, we did another experiment. With a color-printed face photo of the tested person facing Xiaomi Note3, it has not been unlocked successfully. But the cover is used to cover the photo frame to form an irregular border. At this time, Xiaomi Note3 can quickly unlock successfully.
Fig. 8 Color face photos with occluded edges.
Subsequently, we took a photo of the face of the subject with a 5-inch mobile phone and aimed the screen at Xiaomi Note3. We tried many times, but failed to unlock it successfully. After blocking the outer border of the mobile phone with the blocker, Xiaomi Note3 can be quickly unlocked successfully.
Fig. 9 A photo of the face of a mobile phone with its edge blocked.
Through the above three experiments, it is confirmed that the face unlocking function of Xiaomi Note3 can be cracked by using the processed static photos. Next, I hope to verify whether Xiaomi Note3 has video anti-counterfeiting technology. We took a video of the subjects with iPad and played it in a loop. Using Xiaomi Note3 to trigger the face unlocking function on the iPad failed. Next, Xiaomi Note3 can be quickly unlocked successfully after the iPad frame is blocked by the blocker.
Figure 10 iPad face photo with edge occlusion
Conclusion: Through the above experiments, we can crack the living defense technology of Xiaomi Note3 by cutting face photos, blocking the edge of face photos, blocking the edge of mobile phone face photos, and blocking the edge of iPad face video recording. It seems that the face recognition and unlocking security of Xiaomi Note3 needs to be further improved.
Evaluation summary: the technology of letting a hundred flowers blossom has the future face recognition.
According to reports, Xiaomi Note3 adopts the manufacturer's face intelligent recognition scheme and adopts the AI face recognition algorithm, which can effectively prevent others from stealing user equipment by remaking photos, remaking videos and printing photos. However, the actual test results do not seem to be as ideal as expected.
Apple's iphoneX launched one day after Xiaomi Note3 also uses face recognition technology. Its official Face ID technology is based on infrared, binocular camera, 3D structured light face recognition and other technologies, which is different from Xiaomi Note3' s face recognition technology based on big data face modeling. Since iPhone X has not been released in China, I can't actually evaluate it.
With the development of face recognition technology, technologies such as visible light+near infrared, binocular lens and 3D face recognition have been successfully applied to face recognition products in other fields. In the future, this cool "black technology" will be more mature on domestic mobile phones, and it will also replace fingerprint recognition as the standard for smart phones. It is not difficult for domestic mobile phones to catch up with or even surpass Apple. Let's wait and see.