Apple detailed their approach to this last month. They generate synthetic data on their servers and then send it out to iPhones. On the device it compares to real data to see how good the synthetic data is. Then reports back anonymously which synthetic data should be retained or discarded.
https://machinelearning.apple.com/research/differential-privacy-aggregate-trends
Marks
npub136…tjnxf
2025-05-19 13:30:42
in reply to nevent1q…nhha
Author Public Key
npub136jg2fnty2z5vwcnh7p4jpckrs3tk0dpueftgs7mznuuaenjpfps6tjnxfPublished at
2025-05-19 13:30:42Event JSON
{
"id": "113c3fbb1a54cf080ed82879fed03d797ed9e02784fe17929e1ed66d75778484",
"pubkey": "8ea485266b2285463b13bf835907161c22bb3da1e652b443db14f9cee6720a43",
"created_at": 1747661442,
"kind": 1,
"tags": [
[
"e",
"482bb94e27a18c4da605533de7b68c4d3c3ff5502a1e10fc70edc1929c69624d",
"",
"root"
],
[
"p",
"9cb3545c36940d9a2ef86d50d5c7a8fab90310cc898c4344bcfc4c822ff47bca"
],
[
"r",
"https://machinelearning.apple.com/research/differential-privacy-aggregate-trends"
]
],
"content": "Apple detailed their approach to this last month. They generate synthetic data on their servers and then send it out to iPhones. On the device it compares to real data to see how good the synthetic data is. Then reports back anonymously which synthetic data should be retained or discarded. \n\nhttps://machinelearning.apple.com/research/differential-privacy-aggregate-trends",
"sig": "d8964413201b5020118f32c8e5f947464e1ebe02fc65b221a97f5815c714b9d6c023d8db171f734333d1808180179c19bbf377e1fdf57526e17e1403fac5f8ff"
}