Covid In Area 疫地而處

Understand your proximity risk to covid-19 infected areas in Hong Kong

Covid-19 is spreading fast globally taking toll on many lives. The infection is often caused by exposure to the virus for sustained period of time, largely due to coming into contact with the contaminated infected areas. To guard oneself against infection, it is crucial to assess one’s proximity risk with the infected areas in terms of the distance, location noise, and exposure duration. Based on the assessment, one may the take timely and appropriate measures such as health monitoring, voluntary testing, or re-routing to keep the areas at a distance.

To inform citizens, the CHP (Centre for Health Protection) of Hong Kong government has already made available to the public much information on infected areas such as buildings with confirmed cases, and buildings requiring compulsory testing.

We have designed and developed a novel, private and automated mobile app which integrates all the timely open data from CHP and visualized it as an easily accessible heatmap. Users may browse the heatmap to understand where the infected areas are. They may optionally turn on their GPS for the app to compute automatically and locally their proximity risk over time. Our system is based on the following approaches:
Big data mining to assess proximity risk: We integrate the open data sources from CHP and present it as an informative and user-friendly heatmap for users to browse, visualize and understand at a glance where infected areas are. The heatmap is updated continuously according to the government release. If the user turns on GPS, the heatmap can be used to visualize one’s current location with respect to the infection areas. The app also computes locally user proximity risk based on the distance, location noise, dwell time, number and type of infection venues in the neighborhood. The computed risk is charted over a certain past period of time (say, 2-5 days) for personal and private browsing. An overall cumulative risk, indicated in color, is also provided so that the user may take timely and appropriate actions to manage health and plan routes to reduce virus exposure: (1) Red: Overall sustained high proximity level to infected areas. Recommended to monitor health closely and take voluntary testing if possible. (2) Yellow: Medium risk. Be cautious. Take safe paths to reduce risk. (3) Green: Low risk. Stay vigilant.
Machine learning for power-conserving GPS sampling: Users may opt to turn on their GPS to compute and chart proximity risk. To conserve battery power due to GPS, we have devised and implemented a light-weight machine learning approach to adaptively adjust GPS sampling according to user mobility level as inferred from the phone’s IMU (inertial measurement unit). This can greatly reduce the GPS sampling rate, and hence power consumption, by more than 99%.

User privacy is our primary design consideration. Our system is fully distributed without any server collection of any user information. The app achieves full user privacy with the following measures:
Users do not need to register in order to use the app after installation. The app does not ask for any personal information beyond GPS location. It uses no phone data, such as phone number, contact list, gallery, etc.
No any data leaves the phone. All computations are carried out with results presented in local phone only without any data upload.
Users may opt to provide at any time their current GPS locations to the app for proximity risk assessment. The risk level is computed in real time, and once it is computed, the GPS location is immediately discarded without storage.