[ffk]

Been a while since I looked. My understanding is many of these techniques such as SMPC are useful only when operations are linear (eg no exponents or tan functions) and on simple branches. Others like homomorphic encryption require operations on the whole data set.

I think the best best for some of these workflows will be differential privacy on large datasets with strict privacy budgets.

Still highly valuable, but make sure you understand the limitations and risks if you use these techniques.

[jonahbenton]

This piece is mostly about enclaves, which are going to be little-r revolutionary.

Enclaves allow for nearly any kind of processing of an individual's data without making that data directly available to either the algorithm provider or the cloud host. The enclave's public key is provided to the data holder. Their data is encrypted, sent to the enclave, decrypted with the enclave's private key, processed in the enclave, then results are encrypted with the data holder's public key for return to the holder. Other kinds of IO generally not permitted in the enclave, preventing leakage.

A lot of ergonomic details remain to be worked out but the underlying machinery is quite interesting.

[Jeff_Brown]

What questions could never be answered safeguarding privacy? I can think of one -- the taxes paid by the person in the data with the most assets, if you know who the country's richest person is.

But i can't think of any regressions like that.

[sparky_z]

You're assuming that the richest person also pays the most taxes. Not necessarily the case.