KIT researchers warn: New technology recognises people without their own devices based on signals
Researchers at the Karlsruhe Institute of Technology (KIT) can recognise people based solely on WLAN signals. They point out that this poses a significant risk to privacy. The astonishing thing is that people do not need to carry a smartphone or tablet with them to be identified. It is sufficient for Wi-Fi devices in their environment to communicate with each other.
Image like a camera shot
According to the experts, the interaction between the devices creates an image – comparable to a camera shot, but based on radio waves. ‘We observe the propagation of radio waves and can thus generate an image of the environment and of people. This works in a similar way to a normal camera, except that it converts light waves into an image instead of radio waves,’ says KIT researcher Thorsten Strufe.
According to the security expert, it is therefore irrelevant whether someone has a Wi-Fi device with them or not. Even switching it off does not protect you, Strufe clarifies: ‘It is sufficient for other devices in the vicinity to be active.’ And his colleague Julian Todt adds: ‘The technology turns every router into a potential surveillance device.’
Commuting as an everyday risk
People who regularly walk past a café with Wi-Fi, for example, could be identified there unnoticed and later recognised – for example, by government agencies or companies. Although there are simpler methods for intelligence services or cybercriminals to observe people, wireless networks could become a nearly comprehensive surveillance infrastructure, the scientists warn. Wi-Fi networks are everywhere.
Attackers do not need expensive special hardware, the engineers explain. Standard Wi-Fi devices are sufficient. The new method exploits legitimate users who are connected to the Wi-Fi. These users regularly send feedback signals to the router within the network – unencrypted and readable by third parties. This creates images from different angles that can be used to identify individuals.
This takes only a few seconds once the underlying machine learning model has been trained. In a study with 197 test subjects, the team was able to recognise individuals with almost 100 per cent accuracy – regardless of gait or perspective. ‘The technology is powerful, but at the same time poses risks to fundamental rights, especially privacy,’ Strufe concludes.
