Machine intelligence at Dropbox
blogs.dropbox.com
blogs.dropbox.com
Details would be welcome here. From the linked portal and whitepaper I just see a discussion about mostly their operations practices that may offer some protections against inadvertent data disclosure to third parties but nothing about how they preserve any of the users’ privacy from their own servers.
Yeah, well, client-side encryption has not been on the menu for ten years now.
(Mention of which has remarkable anechoic properties.)
2) Differential privacy
3) On device learning
4) Other techniques
Pick any or all of them. Explain what you’re doing and why. Don’t write platitudes about how you’re protecting people’s privacy and have built something privacy preserving if you haven’t used any privacy preserving technologies in your system.