Implementation of contact tracing with the Covid-19 Exposure Notification API by Apple and Google on an ESP32 (with [ESP-IDF](https://docs.espressif.com/projects/esp-idf/en/latest/esp32/index.html)).
More information about the Covid-19 Exposure Notification at [Apple](https://www.apple.com/covid19/contacttracing/) and [Google](https://www.google.com/covid19/exposurenotifications/). This is fully compatible with the official API and is meant for people without smartphone or without access to Apples/Googles implementation.
* send/receive BLE beacons as defined in [Bluetooth® Specification (Apple/Google)](https://blog.google/documents/70/Exposure_Notification_-_Bluetooth_Specification_v1.2.2.pdf) and [Cryptography Specification (Apple/Google)](https://blog.google/documents/69/Exposure_Notification_-_Cryptography_Specification_v1.2.1.pdf)
* BLE privacy (change random MAC address in random interval)
* parsing of Exposure Key export binaries as defined in [Exposure Key export file format and verification](https://developers.google.com/android/exposure-notifications/exposure-key-file-format) (big thanks to [nanopb](https://github.com/nanopb/nanopb) for making this easier than I thought!)
* calculating exposure risks/scores (after adding reported keys and storing exposure information) as defined in [ENExposureConfiguration (Apple)](https://developer.apple.com/documentation/exposurenotification/enexposureconfiguration/calculating_the_exposure_risk_value_in_exposurenotification_version_1)
Due to limited storage, I made some calculations. I have fixed counting of TEKs (14 for two weeks), temporary beacons (1000, longest period for temp. storage is 20 minutes, so recognizing about 1000 different beacons in 20 minutes is possible) and exposure information (choose 500, this is like a limit of infected keys to be met). So the biggest limitation is to store beacons permanently after threshold of 5 minutes. That's what those calculations are for to check, if storage is enough for practical use.
overview of storage in bytes without permanent beacons:
| | size (B) | num | overall (B) |
| :-----------: | -------: | ---: | ----------: |
| TEK | 21 | 14 | 294 |
| Exposure Info | 20 | 500 | 10000 |
| temp. Beacon | 32 | 1000 | 32000 |
Additional 4 bytes counting for every type gives overall 42310B used without perm. beacons.
For now, a partition size of 2494464B will leave 2452154B free for met beacons which leads to a total storage of 76629
beacons. This gives the following table, where I added some lower boundaries to calculate with.
| total beacons | aver. per day | aver. for 10 minute window |
So on average it is possible to meet 38 (24 on a lower boundary) different devices inside of 10 minutes. I have no practical experience/numbers how many beacons are stored on average for a 14-days period in currently running ENA-Apps. But I think regarding the average is calculated for 24h (which is quite unpractical because of sleep and hours without meeting many people), the storage should be enough for the purpose of contact tracing.
[Nanopb](https://github.com/nanopb/nanopb) for reading Protocol Buffers of Exposure Key export. Including already generated Headers from *.proto files.