A cyberattack that uses radio waves to fool image-recognition systems can stop them working. So far, it can disrupt a barcode scanner and alter frames captured by cameras, but there is the potential for it to also work on other types of digital detection systems and make them see things that aren鈥檛 there.
Digital cameras contain sensors that convert light into electrical pulses. Additional electrical signals induced in the circuit going from the sensor, known as post-transducer signals, can give the false impression of actual images. If you can induce these signals from a distance, you can make a sensor see things that aren鈥檛 there.
at the University of Oxford and his colleagues have used radio waves to manipulate a barcode scanner from half a metre away. Similar technology is also being pursued by the US military.
Advertisement
The slight blurring effect on a photo the scanner takes of the barcode is barely visible to humans, but caused the scanner to fail 99 per cent of the time when reading barcodes. This attack simply added 鈥渘oise鈥� to the data the camera was collecting, but K枚hler says his team has already introduced actual shapes in such attacks like readable text (see image above). More sophisticated attacks to fool object-recognition software to make it think something else is there are also feasible, says K枚hler.
鈥淲e are confident that with sufficient knowledge about the target image sensor, such as resolution and sample rate, and with a sufficiently strong attack signal, object-detection systems can be fooled,鈥� says K枚hler.
He notes that fooling a human who might be looking at CCTV footage or through a night-vision scope would be more challenging, but not necessarily impossible.
K枚hler鈥檚 team used a radio-wave transmitter with a limited range, but a more powerful transmitter could inject signals from further away. 鈥淎n attack from tens of metres is possible with reasonably sized hardware,鈥� he says. K枚hler presented the work at the recent 17th ACM ASIA Conference on Computer and Communications Security in Nagasaki, Japan.
The attack works on cameras that rely on technology called charge-coupled device (CCD) image sensors, used in many scientific applications and military surveillance and defence systems, but doesn鈥檛 work on cameras that rely on complementary metal-oxide-semiconductor (CMOS) image sensors, which are now the main type used in consumer devices, such as smartphones.
In a military context, such attacks could create phantom targets or other images to confuse sensors, such as making an autonomous vehicle detect non-existent obstacles in its path to make it stop or change direction.
US Special Operations Command (SOCOM), part of the military, is as part of its .
鈥淗aving previously focused on kinetic solutions, we are now pursuing non-kinetic, innovative ways to win the war without using bullets,鈥� says NGE team leader Kanesha Humose.
Humose was unable to discuss details, but indicated that NGE aims to achieve dramatic new capabilities rather than modest improvements. 鈥淲e are looking to technologies which are more disruptive versus incremental,鈥� she says.
If K枚hler鈥檚 distance estimate is correct, any such weapons would still need to be fairly close to their target to work. This might explain why the technology is of interest to SOCOM, which oversees special forces that carry out clandestine operations behind enemy lines. Humose indicated that research was still in its early stages and they are probably several years away from a fielded system.
Reference:
