Face security recognition, a security technology that was originally only seen in science fiction movies, is now more widely used in real life. But scientists from the University of North Carolina in the United States have reported that the reliability of this technology is worrying. They found that relying on photos generated from Facebook to generate 3D models or 3D printed masks can crack more than 4/5 based The security system of face recognition technology. The surprising conclusion is that scientists at the University of North Carolina participated in the Usenix security conference earlier this month in their paper "Virtual You: Building a Virtual Model from Your Public Photos to Crack Real-Time Face Detection (Virtual U: Defeating Face Liveness Detection by Building Virtual Models from Your Public Photos). This also sounds a wake-up call for the majority of netizens, try not to let their photos appear on the network, otherwise only three low-resolution photos are enough to reproduce your 3D model to fool your authentication system. Researchers call this a "very direct and serious threat" to security. They further believe that "VR-based deceptive attacks fundamentally point to the serious weakness of camera-based authentication systems: systems that rely on color image data and cameras are vulnerable to this, unless combined with other verifiable data sources. Attack in the virtual world." A slide from the Unsenix presentation. All of this was proven at the Usenix conference, where scientists successfully unlocked the facial software system using a digital 3D face model. In addition to this, they also emphasized how easy it is to achieve this through an experiment. There are 20 volunteers (most of whom are security technology researchers who almost never share information on the Internet) as “network stalkersâ€. Researchers found several photos of these people on Facebook, LinkedIn, and similar sites. One participant has only uploaded two photos in the past three years, even though this is enough. These data are then used to generate a 3D model of the face with missing areas, shadows or textures filled by hand. The researchers even added facial movements (frowning, smiling, etc.). These models are so detailed that the 3D faces can even move according to the rotation of the device. “For an observing face authentication system, the depth and motion cues shown are exactly the same as their expected faces,†they said. The researchers used these 3D faces to test five facial authentication systems on the market: Mobius, KeyLemon, TrueKey, BioID, and 1D. The results showed that four systems could be compromised, with success rates at 55% to 85%. And if you can get a specific head shot taken in an indoor environment, all five systems can be broken. "We use social media photos for facial reconstruction primarily to emphasize that a person's appearance is as important to the privacy of the web as other people's information, such as age and location," the researchers concluded. The problem, however, is that in the age of social media, this threat cannot be completely resolved. The only possible solution is to enhance its ability to recognize fraud on the face recognition tool. In particular, the ability to reject faces with low-resolution textures is absolutely necessary, in addition to using light projection patterns, adding infrared-aware sensors, and identifying minor skin tone fluctuations associated with the pulse. Although not tested with 3D printed masks, the University of North Carolina research team said the mask should provide the same results. So for facial recognition software developers, they still have a lot of work to do.
cabinet support
used in kitchen
soft-closing
hot selling in EGYPT and Mid East
Lid Stays,Cabinet Lid Stay,Soft Close Lid Stay,Soft Close Lid Support Wenzhou Zhaoxia Hardware Co.,Ltd , https://www.zhaoxiahardware.com
Scientists discover that face recognition systems can be cracked with 3D printed masks