This article is originally published by AI Frontier. The original link is t.cn/RTXrze4
Edit | Vincent
AI Frontier’s introduction: “A young woman born in the 1990s posted an angry post against the live-streaming platform of the water drip camera, targeting 360; Zhou Hongyi issued an emergency response, emphasizing the black pr encounter. The 360 water Drop incident has raised collective concerns about personal privacy and security. Do we in the smart age need to hide from ubiquitous cameras like in the movie? Can ordinary people’s concerns about privacy be addressed in the face of ever more powerful technological advances? How do we fight this battle for privacy in the smart age?”
The public has been a little anxious lately, a little angry; Recently, an entrepreneur was also anxious and angry.
An article titled “A 1992 female student to Zhou Hongyi: Stop staring at us” has pushed the Internet security company, which has been in the security field for 12 years, to the forefront of the controversy. The specific content must be more or less known to all readers, we simply summarize the girl’s point of view: 360 Water drop camera’s live broadcast function has exposed the privacy of countless people. As soon as this article was published, it was immediately swept in the circle of friends of many netizens, and 360 immediately became the target of public criticism.
Qihoo has responded to the incident, and Zhou Hongyi has also explained the incident on his weibo:
After the incident, the topic of surveillance cameras and privacy on the network caused a heated discussion, and even some netizens said: this incident makes people can not help but think of a few years ago “PRISM gate”, all kinds of secret leaks, but also mostly from that a small camera.
What is the current state of surveillance technology?
In order to make this article easier for readers, we thought it would be worth talking briefly about the technology behind surveillance cameras.
The basic technical discipline of intelligent monitoring is Computer Vision (Computer Vision), referred to as “CV”, which is a science that studies how to teach machines to “see”. Further, it refers to the use of cameras and computers to replace human eyes for target recognition, tracking and measurement of machine Vision, and further graphics processing. Make computer processing into images more suitable for human eyes to observe or transmit to instruments for detection. In the field of monitoring, the most commonly used is face detection and face recognition technology.
Face detection has three key technologies:
- Feature based face detection technology
Face detection is carried out by using color, contour, texture, structure or histogram features.
- Face detection technology based on template matching
Face template is extracted from the database, and then a certain template matching strategy is adopted to capture the face image and extract the picture from the template library to match, by the correlation of the level and matched template size to determine the size of the face and location information.
- Face detection technology based on statistics
By collecting a large number of positive and negative samples of “face” and “non-face” images, using statistical methods to strengthen training of the system, so as to realize the detection and classification of face and non-face patterns.
Based on the three key technologies, monitoring can be carried out through the following four features of face recognition:
- Characterized by distances and ratios between facial points;
- Face image features are extracted with different probabilities of different feature states.
- Face images are regarded as random vectors and different face feature patterns are identified by statistical methods.
- A large number of neural units are used to store and remember the features of face images by association, and face images are recognized accurately according to the probability of different neural units.
Can be so simple to understand: face detection is to judge whether there is a face in a photo; Face recognition is figuring out who a face is.
According to the data collected by AI Front editors, the facial recognition monitoring used in police scenes, such as the subway monitoring, can clearly capture the clear face of almost every passenger getting on and off the bus. If a suspect is to be arrested, it can be said that the arrested person is almost impossible to hide.
Is it possible to avoid all this surveillance?
As noted above, those caught are “almost” inescapable.
“Almost” means it’s still possible to escape.
Here’s what we learned from some of the data we gathered, along with a few brief interviews with technology experts, on whether we can evade today’s smart surveillance:
Face detection has been very mature, there are a lot of detection technology has been added to the gait recognition, at the same time sampling is not only limited to the face, but above the shoulder, in short: it is completely no problem to identify this is an individual.
There are some challenges in identifying the person, such as lighting, expression, occlusion, similar faces, and so on, so that the intelligent surveillance can be fooled.
Not long ago, across the Pond in the United States, a young man challenged the National Security Agency (NSA) to recognize faces:
The exact cause of the incident is not known, nor is it mentioned in my article. All we know is that he chose to paint the features of his face in camouflage to avoid NSA surveillance. According to him, the facial-recognition algorithm pixel computing robot translates his face into a bunch of unremarkable pixels. His face, painted in camouflage, appears in the computer’s view, causing instant confusion.
Although he managed to fool the facial recognition system, he admitted that wearing a big face every day made him stand out more in the crowd, and the psychological side effects made him feel more stressed:
“The look in their eyes made me realize that my strange behavior would cause the public to distrust me. I couldn’t help thinking, if one day I was really sick, really sick, would the people on the street help me? Will the camouflage on my face marginalize me and discredit the public? The camouflage on my face made me uneasy, and I worried that the marks on my face would put me out of place, give the impression that I was playing a prank or a play, and put me at risk without help if I needed it.”
The American boy covered up the basic features of the face, which made it impossible for the surveillance to recognize. The hiding features were one of the reasons for cheating the surveillance, and the fundamental reason was mentioned by several experts and technicians interviewed by us: the size of the target database.
One of the experts interviewed mentioned that the accuracy of face recognition is affected by the size of the target library. For example, if the company clocked in at 1:1000, it would be able to identify the person with a mask. If the face catch bad guys 1:1,000,000,000 comparison, wearing a mask will not work. Given a photo, determine which person is in the library. The accuracy depends on the size of the library. The current popular face payment is actually a form of facial verification, which refers to the fact that a given photo is already known and whether it is the same person as someone in the database. The comparison is 1:1, which is less difficult than face recognition.
Do we need to hide from surveillance?
It is possible to avoid face recognition surveillance. Ordinary people do not commit crimes or break laws. Is it necessary to specifically avoid these ubiquitous smart cameras?
According to AI Front, in fact, more than 90% of the established points cannot be directly used for face recognition, and the resolution of portrait images extracted from the existing surveillance screen is difficult to reach 40*40pixel. Low resolution images will lose a lot of high frequency information compared with high resolution images, and the richness and expression ability of detail information they can provide will be reduced. For the same algorithm model, the lower the image resolution is, the lower the recognition accuracy is. If you want to use cloud center structure to build face recognition, it means a lot of new hd video in the future.
In order to make the accuracy of the face recognition system to meet the requirements of practical application, the distance between the eyes of the face in the picture can reach more than 40 pixels, which has certain requirements for the height and Angle of the camera. Existing security monitoring point of more than 90%, even with 2 million pixels, and even higher resolution, but as a result of building height, Angle of monitoring is not designed for face recognition and construction, still cannot be directly used for facial capture, while those in shopping malls, subway, crossroads and so on population flow huge areas of surveillance camera requires more sophisticated technology.
It is not difficult to see from these data that the general security surveillance cameras have little impact on citizens’ privacy. What we really need to pay attention to are those household webcams, computer cameras and even mobile phone cameras that exist in our daily life. After these devices are hacked by people with ulterior motives, in addition to personal data leakage, private life may even be exposed on the Internet, for the “appreciation” of millions of eyes.
What about the future?
As technology continues to advance, the challenges mentioned above will gradually become less of a problem. At present, some enterprises are exploring infrared + face recognition monitoring, which is said to be a perfect solution to the problem of facial occlusion. In addition, the Face ID Face unlocking technology carried by the popular iPhone X represents the progress of Face recognition technology and the development of technological democratization. Face recognition technology has begun to move towards 3D, although the security still needs to be improved.
We cannot stop the progress of technology, but by the progress of privacy will also become the people focus on the problem, we hope that no matter from the technical or the policy level, to be able to produce an effective plan, protection of the citizens’ basic privacy, “prism”, after all, one is enough, who will also do not want to come up to the second, third, the NTH snowden, Isn’t it?
Follow our wechat account “AI Front “and reply to “AI” in the background to obtain the SERIES of “AI Front “PDF e-books