[16:58 Thu,24.July 2025 by Thomas Richter] |
Security cameras quickly reach their limits when identifying people in poor lighting, concealed people or unfavorable camera angles. Italian researchers at La Sapienza University in Rome have now presented an alternative method that makes it possible to identify and recognize people using only WLAN signals.
;) Recognition via WLAN has several advantages, because compared to security cameras it also works in poor lighting conditions and can even "see" through walls or other obstacles. It also solves the problem of recognizing a person, i.e. ensuring that a person is really known. Visual identification via camera is prone to errors and can be easily deceived, e.g. by sunglasses. How easy or difficult it is to trick identification via WLAN remains to be seen - how can the body signature be fundamentally manipulated? Is a filled backpack enough? The researchers see another advantage in the fact that the raw data generated during person recognition via WLAN is anonymous and therefore useless to an attacker in the event of theft or hacking, for example - without the specifically trained AI model and the associated system architecture, no one can draw conclusions about the recognized people from this data. However, the system can also be used to monitor people if the biometric signature is linked to a real person using other data, for example to track the movements of a customer in a store exactly. ;) To recognize the individual absorption patterns, the researchers use a neural network with a transformer architecture, which analyzes the Channel State Information (CSI) of a WLAN, i.e. data about the waveform of each subcarrier of a WLAN network, which is constantly changing due to the presence and movement of people in a room. The system achieved an accuracy of 95.5% in correctly identifying a person by their biometric WLAN signature. However, a separate WLAN receiver with three antennas is also required in addition to the WLAN router. More than 2 years ago, researchers at Carnegie Mellon University demonstrated how WLAN routers can be used as room radar to reconstruct the postures and positions of several people in a room URL (www.slashcam.de/news/single/KI-verwandelt-WLAN-Router-in-Raum-Radar-17788.html). Another study URL (dl.acm.org/doi/10.1145/3191755) showed that the resolution of such a WLAN radar is even sufficient to recognize gestures of individual people - in this case, the system was able to distinguish between 276 finger and hand gestures of sign language. Together with the new WhoFi algorithm, individual people could thus be identified and their respective movements, including their gestures, could be precisely tracked, recorded and analyzed - even through walls. An analysis of the almost ubiquitous WLAN signals would be sufficient. The possible uses for such methods are diverse - from legitimate applications, for example in the field of security technology such as remote monitoring of your own home or office for illegal intruders, to medical monitoring, where the rooms of a nursing home could be scanned to see if a person has fallen. However, the WhoFi method also harbors considerable potential for misuse. The technology could be used for criminal applications such as espionage or spying on private rooms up to stalking individual people. Here is the pre-print study URL (arxiv.org/html/2507.12869v1) ("WhoFi: Deep Person Re-Identification via Wi-Fi Channel Signal Encoding."). Bild zur Newsmeldung:
 more infos at bei arxiv.orgdeutsche Version dieser Seite: WhoFi - Neue KI kann Personen nur per WLAN-Signal identifizieren |
;)
 |
|
|