cw1/nhs-v-google-apple.md
Harvey Tindall 4aa94ef95b
add nhsx v apple/google api comparison
might be useful for Australia, as they released an app using their own
tech so comparisons could be made to nhsx tech.
2021-10-13 16:23:52 +01:00

2.2 KiB

Original NHSx app vs Post-may Apple/Google API-based app

  • NHSx app builds a database of significant contacts (<2 metres for at least 15 mins) who can be alerted if the user reports symptoms (presumably a positive test, too) (from independent)

    • Users could be notified of false positive if this was found out later
  • NHSx uses a centralised model (FIXME: Find decent source and explanation!)

    • A centralised server figures out the matching process
      • Phone has a unique key thats sent to other devices that come in contact
      • If the user logs positive, device tells server their key is positive
      • Other device asks NHS server if any device keys it's received are positive (here, the server checks for matches)
      • If the key is found to have logged positive, the user is alerted
    • The server knowing if your key has come into contact with someone could pose a risk of tracking by gov/hackers
    • Could make tracking outbreaks on a national scale easier
    • App connected to NHS account
    • Battery-saving features on phones (especially iPhones) could pose a challenge to keeping the service running 24/7 on the device
      • Even if not, Apple claimed NHSx solution would be less power-efficient
  • Apple/Google uses a semi-decentralized model (good ass diagram here)

    • Uses BLE to reduce power usage
    • No NHS account, name or details need to be given
    • When two phones meet, a unique key is created and exhanged between them
    • When one logs a positive, the user agrees to upload the key to a central database
    • All other phones regularly download the central database, and then check (matches are searched for on-device) for any keys they have received from other devices
    • If the key is in the db, the user is alerted, but the central server does not know about the match.

Similar between the two

  • Risk of high false-positives/Self-certification being used maliciously
    • This was more of a concern when the apps were in development, as conveninent testing (LFD) was not easily available so self-cert was the most likely way to log "positive" on the app