Really want boring simple cryptography with few assumptions; used Rust bc strong typing enabled misuse-resistant API design, and performance came for free #realworldcrypto
How do we bootstrap trust in the org key? It helps to be part of a long-trusted publication #realworldcrypto
'Words matter' #realworldcrypto
When sufficiently motivated, users do read the manual #realworldcrypto
Secure elements are great tools...except they don't expose auth attempt counters, nor enforce timeouts (they are reserved for the operating system) #realworldcrypto
RainbowSloth key stretching scheme #realworldcrypto
Q: Diff from SecureDrop? A: Focused on the first contact experience; using Tor is a red flag/signal; we don't support sending files, pivot elsewhere for that #realworldcrypto
Work with actual UX researchers; in-person user testing; #realworldcrypto
Speaking of SecureDrop, next up is 'SecureDrop Next Generation: Lessons from a Decade of Deployment', presented by Rowen Shane, Shannon Veitch #realworldcrypto
Uses Tor, PGP encryption, #realworldcrypt
Deployment/maintenance burden, physical security requirement, better encryption is available #realworldcrypto
What do we want from a next-gen SecureDrop? #realworldcrypto
Don't want to leave extra evidence behind #realworldcrypto
'The real way to use deniability is in combination with other security properties of the system; there are tradeoffs against authentication' #realworldcrypto
Client is the Tor browser; no persistent state; server is 'just' an encrypted message queue; one time keys #realworldcrypto
Key bundles per-journalist #realworldcrypto
Pros, and cons #realworldcrypto
Different organizational profiles motivate different deployment types #realworldcrypto
GitHub - freedomofpress/securedrop-protocol: Research and proof of concept to develop the next SecureDrop with end to end encryption.

Research and proof of concept to develop the next SecureDrop with end to end encryption. - freedomofpress/securedrop-protocol

GitHub
Q: Target a particular journalist, key exhaustion? A: Known, not seen in existing SecureDrop instances, so deferring doing anything for now, but there are approaches #realworldcrypto
Next up, 'At-Compromise Security: The Case for Alert Blindness', presented by Simone Colombo #realworldcrypto
How much are our assumptions and models actually based on the real world? #realworldcrypto
"Cryptography is also a social science unaware of itself." #realworldcrypto
82 abductions related to these protests #realworldcrypto
Vigorous application of ethnography to cryptography #realworldcrypto
'deep hanging out' (technical term) #realworldccrypto
Autumn School · Social Foundations of Cryptography

Adverary goal: intelligence, via torture #realworldcrypto
'raising the alarm' saved victims #realworldcrypto
"The abductors eventually heard the message, which put pressure on them to release the abducted person." #realworldcrypto
At-compromise alert blindness: when they are taken, abducted people can alert their network without the adversary noticing #realworldcrypto
Personal secrets capture the distinction of compromising people or their devices. Multiple personal secrets model that the abducted person knows they have been abducted: they can decide which secret to use. #realworldcrypto
Blindness: 'If I do anything now to upset these guys, I’m dead.' #realworldcrypto
We give up confidentiality for alert: a goal only ethnography could suggest. We give a provably secure construction and we implement it. #realworldcrypto
Ethnography proposes 'exists' statements, not 'for all' statements #realworldcrypto
Coffee time #realworldcrypto
'Defending the Open Internet: TLS, Passkeys, and the Privacy Stakes of Digital Identity', by Christopher Harrell #realworldcrypto
Q: How do I resolve the dilemma that if i work on this, governments will use it to discriminate [against children, women, etc] A: Build primitives that work with trust lists, group of issuers we trust [um we trusted the US until we didn't, oops] #realworldcrypto
Q: Device-bound means i lose my creds A: Cloud-based HSM [narrator: 🙄] #realworldcrypto
Not all have PQ privacy #realworldcrypto
Q: The nice ZKs going to PQ change the computational/privacy properties A: Depending on the construction (BLS vs ML-DSA, relying on the symmetric guts of certain constructions) #realworldcrypto
Next up, 'Zero Knowledge (About) Encryption: On the Security of Cloud-based Password Managers', presented by Matteo Scarlata and Giovanni Torrisi #realworldcrypto
'E2EE' password managers #realworldcrypto
We don't trust the server, should only have access to the vault plaintext if you have access to the master password #realworldcrypto