Snagging an unsecured Wi-Fi signal may occasionally grant an unsolicited peek into someone’s data traffic and computer, but what if that same signal, unsecured or otherwise, could be used to spy on their silhouette’s throw a wall? Worse, what if it could recognize them in a crowd? MIT researchers from the Computer Science and Artificial Intelligence Lab have developed software that intercepts rebounding wireless signals to see behind wall
The technology — called RF-Capture — is precise enough to track the motion of a human hand with reasonable accuracy and, when fed through a machine learning algorithm, can recognize people based on their unique Wi-Fi silhouette.
RF-Capture is the natural evolution of a similar software developed back in 2013, with the added benefit of measuring the signal change reflected off a human body.
How this is accomplished is simple: broadcasted Wi-Fi signals bouncing off of reflective surfaces are intercepted by a capture device, which then applies the RF-Capture software to sort the signals and separate those bouncing off human body parts from other forms interference. The software then consolidates the data snapshots from each body part into a collective silhouette representing the human body.
By combining machine learning with the distinct difference in body metrics of the photographs – height, shoulder width, and more – it becomes possible to train algorithms to spot the subtle difference in persons and recognize who is standing behind the wall. “[W]e use the captured human silhouettes from our reconstruction algorithm [to] train a classifier on these silhouettes which allows us to distinguish between people,” explains Fadel Adi, one of the project’s researchers. “The classifier captures features like height and body builds, which allows us to distinguish between people using RF-Capture.”
How accurate is it? Testing revealed that the system can identify up to 15 people behind a wall with 90 percent accuracy. As far as tracking actual points in real-time, it can trace a person’s moving hand within about an inch, said Adi, suggesting that the technology may have some very interesting real world applications as it improves. For example, the RF-Capture’s ability to ascertain a person’s breathing and heart rate could spark health tracking implements that watch over senior citizens, or it could integrate some sort of gesture control for IoT devices scattered throughout the home.
But, as with all new means of obtaining personal information, the biggest elephant in room is privacy. To prevent abuse, the team insists that blockers must be developed to prevent any solutions piggybacking off of RF-capture from tracking people other than the owners of the device.
Source: Gizmodo
Advertisement
Learn more about Electronic Products Magazine
