Scientists show routers can be used to track your movements through walls

Leveraging a tool called DensePose from Facebook’s AI Lab, the system analyzes invisible Wi-Fi radio signals that permeate a space and creates a visual model. Radio signals broadcast by a Wi-Fi router naturally bounce off objects and people, and the reflected signals carry information that AI can filter to reconstruct body posture and movement. In the same way that LiDAR or echolocation can be used to create a rough three-dimensional representation of rooms or areas, DensePose can produce images of humans and, over time, could be extended to track and understand a person’s movements.

This has already been done. Years ago, MIT researchers built a similar system using RF-Capture technology. Still, the models they presented were not as high-quality as those obtained by the CMU team and the DensePose recreations. The key to DensePose is what the researchers describe as a “deep neural network that maps the phase and amplitude of Wi-Fi signals to UV coordinates in 24 human regions.” With Wi-Fi signals as the only input, this AI model can estimate the pose of multiple subjects.

Although the CMU study has not yet been peer-reviewed, the scientists demonstrate that their DensePose-driven system can monitor human movement inside a building. But there are a few things to keep in mind. To begin, they use a developed neural network model to make estimates and analyze the data. This includes training the model on the layout of rooms and spaces. Although the DensePose GitHub repository is accessible to everyone, the trained model is not. Bad actors should access or develop their own for active surveillance and train extensively using known spatial configurations.

Additionally, the study does not address the technology’s effectiveness in an average residential home. The researchers used relatively inexpensive devices, but they needed multiple access points: the radio waves came from three routers and three receivers. Without a mesh router system, people might only have one router and one or two repeaters placed throughout their home. It’s unclear what impact this would have on the models. Additionally, in the average home, many factors can interfere with Wi-Fi signal strength, which could also be a concern for a system so dependent on reliable readings.

However, larger facilities with multiple network devices (such as hospitals, offices, or commercial buildings) and access to higher-quality signals may be able to provide sufficient information. The study also points out that challenges increase when tracking multiple subjects, so it may be more difficult to track an entire family or groups of people with something like this.

Regardless, research demonstrates that it is possible to locate and track subjects using only Wi-Fi as input. In the study, CMU researchers imagine the technology could be used to monitor the “well-being” of a home’s residents or “identify suspicious behavior.” Questions then arise: who is monitoring, what would be considered “suspicious,” and what actions should be taken when strange behavior is detected?

Going further, if the technology were commercialized in any way and used for market research or data collection, it wouldn’t take much to understand the extent of the privacy issues. Reports show that 80% of American households have a home network router, which means open access to Wi-Fi signals in their home. Wi-Fi imaging could enable passive surveillance in homes and other buildings without physical access or consent. Additionally, a general difference between Wi-Fi and wireless Internet is that Wi-Fi is used to distribute the network in a limited space. This is how Wi-Fi networks remain active when the Internet is down, and in this case, such a system could be re-engineered to work locally without Internet access.

For now, the limitations may hold him back until someone finds alternative solutions. However, it is unclear how long this would take. Future networking technologies, already here and solving the biggest problems of previous Wi-Fi generations, could make consumer routers more powerful and viable for this kind of use.