Anonymous Communication
Format. Break into groups of 4–5. Each group picks one of the two breakouts below (or takes both if time allows). Spend ~5 minutes skimming the prep reads, then ~10 minutes debating. A designated reporter brings the group’s position — and any dissents — back to the full class for a ~3-minute report-back.
Chapter 6.2 makes two claims that live in tension: (1) Tor’s anonymity is only as strong as the infrastructure that hosts its relays and bridges, and (2) the same infrastructure — exit nodes, cloud front-ends for pluggable transports — is precisely where legal and commercial pressure lands. Both breakouts sit on that fault line.
Breakout A: Exit Nodes and the Criminal Safe Harbor
Motion. “Operating a Tor exit relay should carry an explicit criminal safe harbor, analogous to Section 230 for platforms — no country should be able to prosecute an operator for traffic they merely forwarded.”
Prep reads (5–10 min).
Discussion prompts.
- The book notes that Tor has ~2.5M daily users and depends on volunteer relay operators. If exit-node operators can be raided, prosecuted, or bankrupted, what happens to the size and geographic diversity of the exit-relay pool — and therefore to the anonymity of every Tor user in the world?
- The Silk Road prosecution went after Ross Ulbricht as an operator, not against Tor infrastructure. Ulbricht was later pardoned (2025). What signal does that pardon send to (a) prospective dark-web marketplace operators, (b) prospective exit-relay volunteers, (c) prosecutors deciding whether to charge next time?
- Compare an exit-relay operator to (i) a coffee-shop offering open Wi-Fi, (ii) a residential ISP, (iii) a Cloudflare-style CDN, (iv) a Signal server operator. The law treats these very differently. Which analogy is closest, and why does the analogy you pick determine the policy answer?
- A safe harbor for exit operators would also protect people running exits for dark-web marketplaces, CSAM traffic, and ransomware C2. Is there a version of the safe harbor that carves those out without collapsing back into the current situation? Or is the point that you can’t distinguish traffic without destroying the anonymity property?
Bring back. Your group’s draft of a one-sentence safe-harbor clause — the actual text — that you would defend in front of both the EFF and a prosecutor.
Breakout B: Compelled Takedown of Bridges and Pluggable Transports
Motion. “Cloudflare, AWS, and Azure should refuse any government demand to take down Tor bridges, Snowflake broker infrastructure, or domain-fronted circumvention endpoints — even lawful demands from democracies.”
Prep reads (5–10 min).
Discussion prompts.
- The book describes how Google and Amazon disabled domain fronting “under the guise of improving security,” effectively pulling the rug out from under meek. Was that a business decision, a security decision, or a censorship decision? Does it matter which — if the effect on Iranian and Chinese users is the same?
- Snowflake ephemeral proxies run in volunteer browsers; the broker runs on cloud infrastructure. If a government compels the cloud provider to shut down the broker, Snowflake breaks globally — not just in that country. Is that a proportionate exercise of national jurisdiction, or an extraterritorial takedown by design?
- Cloudflare hosts Tor bridges, hosts state-media sites, and has publicly grappled with “who should be allowed on the Internet.” Is there a principled distinction between Cloudflare refusing service to 8chan (which they did) and Cloudflare refusing to host Tor bridges under Russian legal pressure (which they have not)? Or is it just a question of who is asking?
- Iran approved a state-controlled VPN whitelist while blocking Signal. Signal has historically used domain fronting to reach Iranian users. If AWS or Azure caved to a hypothetical Iranian pressure campaign, what’s the fallback? Is there one that scales?
Bring back. A one-page decision tree your group would give to Cloudflare’s trust-and-safety team for handling a government takedown demand aimed at circumvention infrastructure. What are the branches, and what triggers each?
Instructor notes
These breakouts map to the “Pluggable Transports” and “Tracking and Deanonymization” section takeaways in Chapter 6.2, and pull threads from Chapter 6.4 (infrastructure) forward. Breakout A tends to divide cleanly along whether students think Tor’s marginal utility to journalists and dissidents outweighs its marginal utility to criminals — press them to name the specific tradeoff rather than argue in the abstract. Breakout B is where the “infrastructure consolidation” theme from Lecture 1 pays off; the strongest report-backs will connect the dots between commercial cloud policy, national censorship, and the fragility of the Snowflake/obfs4/meek portfolio.