Next up at @enigmaconf: "Platform Data Privacy & Security Strategies for Public Interest Research" by Steven Buccini
#enigma2020
-
-
So the lawyers were not comfortable with differential privacy for this purpose let's see how this looks in practice: edge cases make things hard. So you throw away thresholds and throw away things outside. Messes with dataset and sometimes humans need to help set thresholds
Prikaži ovu nit -
Also: not enough engineers working on this to work fast enough. Also it's *reaaaaally* hard
Prikaži ovu nit -
We need HIPAA for public interest research [ ... this is a bold contention given how HIPAA works. It's ... not my favorite, but lets see ]
Prikaži ovu nit -
Need: * safe harbors if orgs are acting in the public interest for research (not as much protection for research etc) * accountability for situations like Cambridge Analytica * regulate the process rather than the data to be more flexible... like insider trading rules
Prikaži ovu nit -
* trust but verify. academics can't tell whether it's a representative data set from a company. Needs a way to verify that they're getting access to all the data they need to do this research in the public interest
Prikaži ovu nit -
Public Interest Research Alliance was created to bring more voices into the room. Social Science One was between FB and Harvard/Stanford. So get involved in PIRA!
Prikaži ovu nit -
CCPA is going to spur privacy regulation, so good time to work on this [ ... yes, please, federal privacy regs but I am not optimistic PLEASE GOOD FEDERAL PRIVACY REGULATION NOW ]
Prikaži ovu nit -
Q: when you said you could anonymize by hashing. How do you avoid application risk like with the Netflix data set? [ THANK YOU CALVIN ] A: I'm not a crypto expert. Not trying to focus on the methodolgy. It's not useful for public interest research because it obliterates trends
Prikaži ovu nit -
Q: how can this paradigm spread to places where companies have been asked to share data with governments e.g. LA wanting data from scooter/ride-sharing apps? A: municipalities are more nimble so a good place to start
Prikaži ovu nit -
Q: you say this research is in the public interest. who decides this? A: good question. open for debate, but I think that may be better to frame it as a way to have system of checks/balances over the claims that system operators are making
Prikaži ovu nit -
Q: want to thank [other people sorry
@ljean didn't catch their names ]. Can you include consent as a way to get detailed results and ways to compared with the aggregate by using more nuanced, advanced methodsPrikaži ovu nit -
A: not a cryptographer. Regulators in Europe seem really frustrated because they think they made an academic carveout already and they don't like the platforms making legal arguments/excuses Want to be ready when the regulators move forward and want to force the platforms
Prikaži ovu nit
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.