The full set of documents for DP3T is at https://github.com/DP-3T/documents. Please refer to the technical documents and whitepapers for the descriptions of the implementation.

This repository implements both proximity tracing designs from the white-paper. In both designs, smartphones broadcast ephemeral Bluetooth identifiers (EphIDs), and store the identifiers received from other devices. The designs differ in how they generate these EphIDs and how they do proximity tracing:

• Low-cost design. In this design, EphIDs are generated from a rotating seed SKt for efficiency. When an app owner is diagnosed with SARS-CoV-2, their app uploads the corresponding seed SKt. Other apps use this seed to regenerate the ephemeral Bluetooth identifiers, and check if they have seen an identifier corresponding to an infected person.
• Unlinkable design. In this design, EphIDs are generated independently from random seeds. When an app owner is diagnosed with SARS-CoV-2, their app uploads the corresponding seeds for each EphID broadcast in a compact representation. Other apps use this compact representation to check if they have seen an identifier corresponding to an infected person.

Difference with respect to the Andoid/iOS SDKs and backend implementations. The DP-3T project has also published Android and iOS implementations and a backend SDK for use in production apps and backends. This reference implementation serves to provide a simple implementation that complements the white paper. We try to avoid differences between this Python reference implementations and the other implementations, but small differences are unavoidable due to the speed at which this project progresses.

Example implementations of App and Backend. We plan to soon extend this repository with example implementations of the App and Backend server. The App implementation will show how smartphones interact with the backend server's API to upload tracing information, to retrieve tracing information, and to do contact tracing.

This reference implementation is deliberately simple. It aims to provide a clear and simple implementation of the cryptographic concepts from the whitepaper and to show how these tie together.

The package dp3t.config contains global configuration parameters shared between all designs. The package dp3t.protocols contains the reference implementations lowcost and unlinkable for the low-cost and unlinkable designs. These files follow a similar structure:

1. They introduce design-specific parameters such as the length of a batch.
2. They define utility functions to convert time to internal representations
3. They define simple cryptographic functions used to generate seeds and EphIDs
4. The class TracingDataBatch represents the information downloaded from the server. Both designs use the same interface for this class.
5. The class ContactTracer ties these functions together and sketches the core contact tracing functionality that would be used by a smartphone contact tracing app. The class takes care of rotating keys (next_day), generating EphIDs to broadcast, to output EphID given a specific time (get_ephid_for_time), to process an observation (add_obseration), to output tracing information (get_tracing_information), and to process a batch of tracing information to determine the number of contacts with infected people (matches_with_batch). Both designs use the same interface for this class

This code deliberately does not implement any interactions with (simulated) Bluetooth devices or backend services.

You'll need to install the project. For example as follows:

python3 -m venv venv
source venv/bin/activate
pip3 install -e "."

After installing the project you can run the examples:

examples/run_lowcost.py
examples/run_unlinkable.py

After installing the project you can obtain test vectors by running:

utils/testvectors_lowcost.py
utils/testvectors_unlinkable.py

To run the tests, install pytest (e.g., pip install pytest) and then run pytest. You may need to reload the venv (deactivate followed by source venv/bin/ativate) to ensure that the paths are picked up correctly.

Please install the proper pre-commit-hooks so that the files stay formatted:

pip install pre-commit
pre-commit install

This code is licensed under the Apache 2.0 license, as found in the LICENSE file.