Using Wi-Fi Infrastructure to Predict Contacts During Pandemics

Wi-Fi log data can be used as an alternative to Bluetooth and other protocols to detect possible exposure to an infected person without requiring installation of specialized apps or changes to infrastructure. This paper summarizes examination of five increasingly complex approaches to predict individuals’ locations indoors and consequently predict exposure to infections: (N1) Common Wi-Fi access point; (N2) In the building at the same time; (PI) Location predicted from Wi-Fi coverage; (P2) Location predicted from Wi-Fi coverage and previous location; and (P3) Movement predicted from floorplan graph. Data for 12 study participants completing 158 simulated scenarios were collected. The data were further used to create synthetic exposure datasets with up to 1000 people in a building over an 8-hour period. Accuracy of five algorithms applied to predicting locations and contacts was examined. The best results were obtained from method P3 that achieved: recall 0.898 and confidence 0.24 in predicting locations, and AUC 0.86 in predicting contacts. These results indicated that for environments with enterprise Wi-Fi infrastructure through a building, predicting movement using Wi-Fi log data along with known floorplans can accurately predict exposures.