Facial surveillance and recognition systems have evolved considerably, but so have methods to deceive this kind of software. According to cloud computing security services experts, a group of IBM researchers have developed a T-shirt stamped with a kaleidoscopic patch that makes the carrier undetectable for facial recognition systems with artificial intelligence. This T-shirt is a new option of what experts call “adversary items”, which are physical objects designed to counteract some technological implementation, such as surveillance systems.
These objects alter the functioning of neural networks used for object detection. Under normal conditions a neural network recognizes an individual or object which is within its visual field and assigns it a recognition flag.
The group of researchers found the boundary points of a neural network, which are the thresholds within which it is decided whether or not something is a particular object. Thanks to this, experts were able to create a design capable of confusing the network classification system with artificial intelligence.
Experts conducted the tests using two neural object recognition networks (YOLOv2 and Faster R-CNN). As a result, cloud computing security services experts identified areas of the body where it was possible to add pixels to confuse the artificial intelligence system, making the user virtually invisible to surveillance cameras.
Previous research has already sought the right object design to deceive these systems. A couple of years ago, a Carnegie Mellon team created glasses capable of bypassing the most sophisticated surveillance systems to prevent the wearer from being detected. Other research has focused on traffic monitoring systems and speed limits.
However, cloud computing security services experts point out that previous attempts have been tested only on static material, as getting this to work with real surveillance systems is much more complex: “Going completely unnoticed for surveillance systems during a full video sequence is the real challenge,” says Battista Biggio, a researcher at the University of Cagliari.
The creators of this T-shirt claim to be the first to overcome these challenges and evade detection even in a complete video sequence. According to your report, this is due to the use of a “transformer”, a method of measuring how a T-shirt moves and then adding those patterns to the design.
Despite the success achieved during testing, the International Institute of Cyber Security (IICS) anticipates that it will be difficult for these kinds of objects to be used in reality with the same results, as these systems are constantly updated and, depending on how a pattern of shapes and colors is perceived, multiple failures can occur.
He is a well-known expert in mobile security and malware analysis. He studied Computer Science at NYU and started working as a cyber security analyst in 2003. He is actively working as an anti-malware expert. He also worked for security companies like Kaspersky Lab. His everyday job includes researching about new malware and cyber security incidents. Also he has deep level of knowledge in mobile security and mobile vulnerabilities.