The Problem: Not knowing your level of exposure to Covid-19
WHO's aim is to test as many people as possible in order to slow down the pandemic. WHO has sent approximately 1.5 million test kits to 120 countries to meet global demand and the agency is working with companies to make them available to those in need, however, there is a significant shortage of testing kits and resources. Hence, there is an urgent need to identify and assess individuals who have been at the greatest risk of infection to get the best use of available testing resources, and to minimise the infection rate.
Prior to being tested a member of the public has no means of identifying when or where they are exposing themselves to potential risk of infection. We want to make citizens and residents aware of the potential risk of Covid-19 infection through their movements or when taking a journey. As people are the primary means of transmission, we need to be able to trace the movements of an infected person in the community in order to determine whether they have come into close contact with any other person. However, this needs to be done in a secure and privacy preserving way. We do this by using a mobile phone as a proxy for a person, as its movements will directly correlate to the movements of its owner.
The answer to 3 simple yet critical questions will assist an individual assess their exposure to Covid-19:
- Where they have been in the last 14 days?
- Where they currently are?
- Where they would like to go in the immediate future?
Managing the risk identified from the answers to these 3 questions will help everybody reduce their exposure to this virus and ultimately defeat it.
Can Technology Solve this Problem? Yes, by identifying a location in real-time where there is a high risk of infection for an individual. By using the location history of a mobile device in order to identify areas of Covid-19 infection so that a person has an easy-to-use means of assessing their personal risk of infection and do so in a privacy compliant way.
A citizen's problem is that they do not know when and where they may have crossed paths with an infected Covid-19 person. Using COVIDRA, a person’s connectivity history to the MNO cell tower network can be retrieved and cross referenced with another person’s movements, a score will be computed to enable an assessment of infection risk. This information needs to be fluid, dynamic and time relevant to facilitate accurate risk assessment to inform better decision making. This solution provides a foundation for a number of future enhancements that are already under discussion.
The goal is to provide citizens and residents with a risk indicator to assist them in assessing the infection risk exposure because of their past, present or future movements.
Almost all citizens have smartphones. The approach is for a user to know whether their smartphone was, is or will be close to an infected phone i.e. the smartphone of another user who has become infected (see Step 1 below).
The solution is based on the principle that mobile network operators (MNO's) know the location of every active smartphone and its movements between mobile cells, as the phone connects and disconnects. In urban environments, with a high emitter and base station density, it can be possible to identify where the phone is within a cell using directional antennae and also triangulating between cells. The resulting resolution varies depending on the cell sizes but is normally measured in tens of metres. The triangulations are mapped to a set of small squares, each about 3m x 3m, which has a highly efficient global Unique ID. We map these for time and distance:
Time. The triangulations are mapped for cell velocity in km/h; the area of the cell is transformed to a circle with a diameter, which is divided by the time spent in the cell to give the cell velocity. This is done in a way that includes the dwell time in the cell. For airborne infections to occur, the medical assessment is that 15 minutes close proximity is required. For contact surface infections, which are highly dependent on the surface type and temperatures, the time can vary from 2 hours to 2 days.
Distance. We can map to the same cell (high risk) or adjacent cells (lower risk)
Vulnerability or risk thresholds or filters can be applied to the cell velocity, dwell time and distance to give an infection risk score, which could be presented in simple Red, Amber, Green indicators, or in more detailed formats if required. The solution then uses operational privacy preserving Secure Multiparty Computation (SMPC) to make comparisons of two sets of data:
Infected squares. The infection status of each square is calculated every hour or 15 minutes for the previous two weeks (or whatever is decided). This gives 336 time slots of one hour or 1344 slots of 15 minutes each. A Bloom filter is applied to represent the infected squares for each one of the 336 time slots resulting in a Bloom filter bank (see Step 4 in Figure 2). All of this happens within the MNO or internet application provider.
Non-infected squares. The individual user's history of the squares that they have visited during the previous two weeks or the set of squares in their current or intended future location is also transformed resulting in a Bloom filter query.
The Bloom filter query is compared against each filter in the bloom filter bank using SMPC. The results are presented back to the user and, under controlled circumstances (consent or public safety driven) to the user’s MNO and/or the authorities as legally required. This provides the user with answers to the three questions and therefore the ability to assess their exposure to Covid-19.
We are currently applying for funding and we have engaged an independent person to undertake a DPIA. We will fully comply with GDPR and all relevant legislation.