I listened with interest to yesterday’s parliamentary committee on the proposed NHSX contact tracing app, which is being trialled on the Isle of Wight from today. You can see the recording here.
Much of the discussion concerned the decision to follow a centralised approach, in contrast to several other countries such as Germany, Switzerland and Ireland. Two key concerns were raised:
1. Can a centralised system be privacy respecting?
Of course the answer to this question is yes, but it depends on how data is collected and stored. Cryptographic techniques such as differential privacy are designed to allow data to be de-indentified so that is can be analysed anonymously (e.g. for medical research) for example, although there was no suggestion that NHSX is actually doing this.
The precise details of the NHSX app are not clear at this stage but it seems that the approach will involve identifiers being shared between mobile devices when they come into close proximity. These identifiers will then be uploaded to a central service to support studying the epidemiology of COVID-19 and to facilitate notifying people who may be at risk, having been in close proximity to an infected person. Whilst the stated intention is for those identifiers to be anonymous, the parliamentary debate clearly showed there a number of ways that the identifiers could become more identifiable over time. Because the identifiers are persistent they are likely to only be pseudonymous at best.
By way of contrast, a large team of academics has developed an approach called DP-3T, which apparently has influenced designs in Germany and elsewhere. It uses ephemeral (short-lived) identifiers. The approach is not fully decentralised however. When a user reports that they have COVID-19 symptoms, the list of ephemeral identifiers that user’s device has received, when coming into close proximity to other devices, is shared via a centralised service. In fact, they are broadcast to every device in the system so that risk decisioning is made at the edges not in the middle. This means that no central database of identifiers is needed (but presumably there will be database of registered devices).
It also means there will be less scope for epidemiological research.
All of this is way beyond the understanding of most people, including those tasked with providing parliamentary scrutiny. So how can the average person on the street or the average peer in Westminster be confident in the NHSX app? Well apparently the NHSX app is going to be open sourced and that probably is going to be our greatest protection. That will mean you won’t need to rely on what NHSX says but inevitably there will be universities, hackers, enthusiasts and others lining up to pick it apart.
2. Can a centralised system interoperate with the decentralised systems in other countries to allow cross border contact tracing?
It seems to us that whether a system is centralised or not is a gross simplification of the potential interoperability issues. True, the primary issue does seem to be the way that identifiers are generated, shared and used in risk decisioning. For cross border contact tracing to be possible there will need to be alignment on a whole range of other things including technical standards, legal requirements and perhaps even, dare I say it, liability. Of course, if the DP-3T model is adopted by many countries then it could become the de facto standard, in which case that could leave the NHSX app isolated.
Will the NHSX app be an effective tool to help us get back to normal? This will depend entirely on how widely it is adopted, which in turn will require people to see that the benefits outweigh the costs. That’s a value exchange calculation that most people will not be able to make. How can they make a value judgment on the potential risks to their civil liberties of such a system? The average user is probably more likely to notice the impact on their phone’s battery life or when their Bluetooth headphones stop working.