Undoubtably, there has been a big debate in the last couple of months around the topic of contact tracing apps. This debate focuses mainly on the choice of either centralized or decentralized architecture, in order to minimize the processing of personal data and make sure that privacy of participants is protected.

The teams still remaining in the PEPP-PT consortium are perhaps the main representatives of the centralized approach, having published the specifications of ROBERT on April 18, and of NTK the day after. On the other side, the teams in the DP3T consortium criticize centralized systems, pointing out that they can be turned into mass surveillance instruments of governments, and they emphasize on the development of the decentralized model DP3T. However, there have been several studies criticizing the privacy protection offered by decentralized models as well. In the decentralized approaches we should also include the protocol from Apple and Google, which will be soon in the OS of our mobile phones.

Understandably, all these discussions have created a lot of confusion to non-experts. In an effort to clarify some of the issues, we contacted Dr. Apostolos Pyrgelis*, Post-Doctoral researcher in EPFL and member of the DP3T team. In this article we publish his answers on our questions around the DP3T protocol, focusing mainly on how close we are to real-life deployment and the Greek reality.  We would like to thank Dr. Apostolos Pyrgelis for his availability.

Also, we would like to thank our member Dr. Ioannis Krontiris** for reaching out to Dr. Apostolos Pyrgelis and the DP3T team and for facilitating this interview. We will soon publish this interview translated in Greek, as well.

1) What is the relation between Google/Apple platform (GACT API) and DP-3T? Is the plan to deploy DP-3T on top of the GACT platform or does it have its own independent access to the OS (e.g. Bluetooth hardware, storage of keys, etc.)?

The DP-3T project is formed by a group of international researchers whose interest is to ensure that proximity tracing technologies will not violate human rights key to our democratic society. It started independently of Google and Apple and remains independent of them. The DP-3T project is constantly making new proposals and is publishing positions to inform the discussion around proximity tracing. These positions may be different from these companies’ strategies or points of view.

The Google/Apple joint system design (i.e., the GACT API) aims at enabling interoperability for decentralized proximity tracing applications across iOS and Android mobile devices. This is particularly important for the success of contact tracing apps. To this end, the plan is to deploy the DP-3T protocol on top of the GACT platform and the DP-3T project has access to a code base for iOS and Android that is functional prior to the corresponding OS upgrades. Currently, DP-3T project members are working in close collaboration with Google/Apple engineers to provide open-source support for their API, since we expect that the majority of national applications will be built on top of it.

2) Broadcasting continuously Bluetooth beacons can be used to track people around, since one could try to link these messages together and create traces. What countermeasures do you take against this?

In the DP-3T system, the users’ mobile devices broadcast to their vicinity (random looking) ephemeral identifiers via Bluetooth Low Energy (BLE). To prevent user tracking via these broadcasts, the ephemeral identifiers are changing regularly (e.g., every 15 min). We here note that this is a shared feature among DP-3T and the GACT platform.

3) What is your experience and lessons learnt from testing the app with real users? How many manual tracers did you use for these tests?

We here clarify that we did not perform any testing of the application with real users or contact tracers. We only performed field experiments in known scenarios for which we could collect ​ground truth​ that would enable us to evaluate the accuracy of using BLE beacons for distance estimation among individuals, in various settings. A brief overview of these experiments can be found on the following ​video​. We are currently processing the results of the field experiments aiming to identify the appropriate configurations and parameters for reliable distance estimation using BLE.

4) What would cause a false positive or a false negative in the DP-3T system? Consider for example thousands of people stuck in traffic in a busy city like Athens. That means I am in my car stopped for several minutes next to someone in their own car who’s infected. What existing measures are being considered to mitigate this class of problems?

First, we remind that ​traditional, person-based​ contact tracing has a lot of false positives since the majority of users that are exposed to infected others, do not present symptoms and do not get themselves infected. Similarly, it also has false negatives since infected users are unable to recall all the people that they met with in the recent past or identify strangers that they encountered in a bus, a shop, etc.

It is important to distinguish the above false positives/negatives from those that are related to contact discovery, i.e., the fact that two users were exposed to each other in close distance and for a specific amount of time (as defined by the public health authorities), when it comes to ​digital​ contact tracing. In the DP-3T system, the contact discovery process is realized via transmissions over Bluetooth, whose wireless broadcast nature is inherently affected by factors such as physical objects, radio interference, weather conditions, etc. This might lead to contact discovery false positives, e.g., if a contact is registered even though there is clear physical separation, such as a wall, between the users, and false negatives, e.g., if an actual contact is missed due to radio interference. The DP-3T team is currently performing extensive measurements to better understand the performance of Bluetooth communications for distance estimation in various settings and parameterize the application in a very conservative manner such that false positives/negatives are limited. We have not explicitly tested the traffic jam scenario, but to account for such situations the app will allow the users themselves to temporarily disable the contact discovery process.

5) Which factors are affected while deployment of DP-3T is scaled up? Would the protocol scale to the magnitude of millions of users?

The DP-3T protocol is designed in such a way that it easily scales to countries with millions of users without compromising their privacy. User devices need to download from the backend server minimal information per day (a few MBs) — which makes DP-3T also scalable to countries with poor broadband — and require very little time (a few secs) to generate their ephemeral keys and compute the infection risk of their owner.

6) Have you been approached by the Greek authorities regarding deployment of DP-3T in Greece? Do you have any indication in which direction Greece wants to take with respect to contact tracing apps?

We have not been approached by the Greek authorities regarding deployment of DP-3T in their country. As such, we do not have any information about the Greek plans with respect to contact tracing applications.

7) Assuming that Greece opts in for a decentralised solution in the future (DP-3T or other), what information would Greece have to share with other countries regarding visitors and tourists in order to achieve interoperability between decentralised solutions across countries? Would that reveal travel plans of people back to their homelands?

Interoperability of contact tracing systems across countries is a very important factor for their success — especially, in cases of free movement, such as the EU, where people travel daily to other countries for business, leisure, etc. The DP-3T project envisions interoperability between decentralized solutions across countries and is currently collaborating with designers and engineers from various countries to address its technical challenges. In one of the proposed ​interoperability solutions​, users would have to configure their application to receive notifications from the countries that they travel into. Moreover, the homeland backend servers of the infected users would have to forward the relevant data to the backends of other countries that these users have visited. While this would reveal information about users’ travel patterns to their homelands, we believe that this is acceptable for the international success of contact tracing.

8) A major issue for contact tracing apps is persuading people to actually use them. Do you have any indication what is the minimum necessary penetration of the app to the population in order for the app to be effective? Is a high degree of case identification within a population required and does this translate to widespread testing?

Indeed, the success of contact tracing apps depends on users’ adoption and this is why we believe that it is of paramount importance to ensure them that their privacy is protected. While such a large scale deployment has never been performed before it is not clear what is the minimum necessary penetration of the app to the population for it to be successful. However, epidemiologists believe that any percentage of app usage will contribute to the pandemic mitigation efforts. To this end, the DP-3T project is hopeful that the app will have impact for “proximity communities”, e.g., commuters, co-workers, students, that have a suitable density of deployment. Finally, we remark that contact tracing apps should be complementary (and absolutely not a replacement) to traditional interview-based contact tracing and should be combined with public health infection testing policies. What really matters, at the end of the day, is to bring and maintain the virus transmission rate below 1.

* Dr. Apostolos Pyrgelis is a Post-Doctoral researcher at the Laboratory for Data Security of École Polytechnique Fédérale de Lausanne. His research interests include privacy-enhancing technologies and applied cryptography, and enjoys studying problems at the intersection of big data analytics and security or privacy. He received his PhD from University College London and his BSc and MSc from the University of Patras in Greece.

**Dr. Ioannis Krontiris holds a Ph.D. Degree in Computer Science from University of Mannheim in Germany, and a M.Sc. Degree in Information Technology from Carnegie Mellon University in USA, while he is also a graduate from the School of Electrical and Computer Engineering of the Technical University of Crete. He is currently working as a Privacy Engineer at the European Research Center of Huawei in Munich, Germany.