== Getting big companies to run onion services ==
Alec Muffett Levers for facebook Genuine name policy How do we make FB authenticity (give them fb without a hop from the exit node) Website runs faster User is happier More secure More private Almost free Internal postings Round out with 1:1 data collection/conversations
Tool: EOTK: https://github.com/alecmuffett/eotk
Enterprise Onion Toolkit
Alec Muffett's role when initiated: Data privacy groupWhat business/security objections did you run into / what response? How many people come to fb over Tor: 425-500k/mo => surprise Exceptionally motivated / non-robots Went to Community Ops. "Look for this. Pick a sample. Evaluate for evil." Data driven response. Output <10% evil. Higher/lower than normal? Slightly more than general sample. How Hack code Bi-directional translation Get powerful people on board Work with spam fighting teams to ensure you're not overrun by spammers when you launch One-time verification/account Anticipated objections: did stakeholder and impact analysis How to get PR on board? Enabling those who already access us via Tor to do it better, faster, more securely Non-political arguments (e.g., NOT about censorship evasion) Less limiting (video, etc.) Negligible running cost Best alternative for Tor support at FB If facebook ran relays, fb is trying to deanonymize tor to expose them to the NSA If fb funded tor, it would be perceived as buying tor off FB site on Tor = nothing to object to
Baseline what you need Problem statement, awareness Usage metrics External good PR case
Frame as experiment Launch it as we're experimenting: trying to accomplish xxx benefits De-escalates failure is shame: flaky R&D team
Have other companies come to you about how do we do this? Yes, many Tries to help, sometimes works (e.g., propublica)
Need to at least create integration tests to make sure it won't break Negligible running cost
Contributions to external OS projects might be a legal battle: so modifying and distributing becomes a different thing.
Initially, service will be fragile for unpredictable reasons: constant monitoring (chron jobs) required Redundancy Have powerful friends on the inside
Do the simplest, vanilla thing and then iterate on it.
Business cases Libraries Health services Government services Retail
How/how long to come up with facebookcorewwwi.com just grabbed a couple of clusters that were being underutilized having matt jones (-type guy) on your side helps: fb engineer, well respected, gravitas/seriousness