How one steel manufacturer is bringing facial recognition access control to the factory gate
When Posco ICT layered Alchera’s anti-spoofing biometric technology on top of its facial recognition system, the company brought next-generation manpower management to the world’s fourth-largest steelmaker.
How do plant operators effectively secure access to a factory that employs thousands of workers? That’s the challenge faced by the IT and engineering subsidiary of POSCO.
POSCO is a global leader in steel production. Based in South Korea, the company has been innovating world-class smart factory practices since 2015. Innovation, bundled into an agile management system called PosFrame and run by its subsidiary POSCO ICT, aims to find efficiencies by collecting on-site data, conducting predictive analyses, and optimizing processes through AI technologies.
Smart factory technology produces benefits. One year cost savings associated with these changes at a single POSCO mill were $14 million USD.
It also creates manpower risks. The tilt towards smart factories magnifies the need for skilled workers familiar with the system’s operations and vulnerabilities. At POSCO, every worker receives technical training to work effectively in a data-driven environment. In addition, the workforce enjoys a measure of problem-solving autonomy. Field engineers, for instance, can modify automated processes that directly impact their immediate work scope as well as downstream and upstream processes.
System knowledge and access expose the company to industrial espionage, or worse. Biometric technology like facial recognition can verify employee status at plant entrances, but it can be exploited by spoofing attacks.
Alchera was hired by POSCO ICT to eliminate that risk with an anti-spoofing solution.
POSCO ICT selected Alchera as the prime vendor for two reasons: ease of deployment and accuracy. Unlike some systems, Alchera technology works with 2D (e.g. webcams, standard smartphones) and 3D cameras (e.g. special cameras with ToF (time-of-flight) sensors that can detect the depth of field). That means the anti-spoofing solution could be deployed on the existing network hardware.
To accurately detect fraud at factory gates, the system runs a liveliness check. A person walks up to a facial recognition-enabled camera. The camera takes a picture that is processed by Alchera software and then compared to the images in a database of employee photographs. The response is simply to accept, reject, or confirm.
A confirmation result means more information is required. The person in front of the camera is instructed to look up and then look down. Those head poses are photographed by the camera and run through the database.
Because Alchera’s software can create a 3D image from a 2D picture, it is possible to accurately compare the live image of a person’s tilted head with the picture in the employee database (with a standard look straight ahead pose). The result is a match or no match response. If the person in front of the camera does not comply with the instructions (e.g. because they are holding a photograph or video) the request to enter is rejected.
Powered by Alchera, POSCO ICT deployed anti-spoofing technology in 2018 at the POSCO steelworks mill in Pohang, South Korea.
After scanning a live image, the system typically requires less than 2 seconds to generate a response. Pilot studies indicate the false positive rate is less than 1% (i.e. the system incorrectly rejects a valid employee).
“Nowadays the question is not any more whether or not biometrics can be copied or forged, but rather to what extent systems are robust to these attacks and if they incorporate the necessary countermeasures to detect them.” – “Biometric Anti-spoofing Methods: A Survey in Face Recognition,” IEEE Access · January 2014
What is sppofing?
Spoofing is the ability to fool a biometric scanner (e.g. facial recognition technology) into identifying an illegitimate user as the legitimate user by showing the sensor a fake version of the original. After the system accepts the spoofer, they are free to enter.
There are two types of spoofing attacks. In a 2D attack, someone aims to gain entry by displaying a photograph or video in front of the facial recognition camera. In a 3D attack, someone aims to gain entry by wearing a mask, or displaying a sculpture or robot with makeup.
We’re a technology startup from South Korea. Our domain is artificial intelligence (AI) and augmented reality (AR). Supported by a team of experts in computer vision, deep learning, graphics, data science, and software engineering, we solve problems.
Since 2016, Alchera has empowered businesses to launch products, improve efficiencies, and bring the power of visual AI to their organization.