Per @RachelTobac: 75% of Twitter 2FA users are using SMS-based authentication. In theory those users could switch to authenticator apps (or pay 😂) but they probably won’t.
People keep saying things like “but authenticator apps will still be free and those won’t require you to pay, plus they’re more secure.” That’s true! But also completely misunderstands what’s about to happen.
What sets SMS 2FA apart is that it’s almost “free” from a user-effort perspective. If you own a phone, the feature is already built-in and enabled. Setup is nearly effortless. Backup is taken care of. Unfortunately none of the same things are true for HOTP/authenticator apps.
The cognitive overhead of installing an authenticator app (and then worrying about what happens when you lose your phone) is absolutely ridiculous. The overall experience is just stunningly bad, given that it’s one of the best defenses we have.
Free one-time code authenticators *should* be built into every phone. They *should* be enabled on the default keyboard. They *should* be securely backed up to an end-to-end encrypted account. If Google/Apple did this, adoption would be high.
Instead we have this ecosystem of crappy apps that you have to install manually. Some have weird cloud backup built in, of unknown security level. Some require you to back up manually with a QR code (ugh, Google Authenticator). It’s such garbage.
I learned recently that iOS has authenticator capability built in, and it will even back up to iCloud Keychain (using end-to-end encryption)!
All you have to do is navigate to a hidden submenu buried under “Settings”. It really sucks. But at least it’s better than Android.
I know companies like Apple and Google have all these great ideas like “let’s eliminate passwords using cool new ideas, eg ‘passkeys’”. Maybe that’ll work out. I hope so! I hate that these companies are slow-rolling security *today* so things can be “perfect” tomorrow.
Anyway. I’m certainly no expert on usable security. My only superpower is that *I am extremely lazy.*
SMS 2FA sucks sucks sucks, but I empathize with the Twitter users who chose it — and I understand why the alternative will be “no 2FA.” Too bad for Twitter.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
One of the things I’m trying to explain to my blockchains class this week is that “algorithmic stablecoins” and “backed stablecoins” are fundamentally different, and the fact that they both have “stable” in their name is confusing.
One type of system creates a direct bridge to the traditional banking system. If you have that bridge (and it works, that is: allows someone to deposit and withdraw) you don’t need any other infrastructure, like centralized exchanges etc.
The other doesn’t create a bridge to anywhere. It’s trying to extract a feature (stability) from the fact that other bridges to traditional banking exist elsewhere. The result is the same, obviously the tech is different. But the “system” and its implications are very different.
I got into a discussion with @OrinKerr the other day about the recent NYT op-ed on Signal and whether “metadata-resistant communications” is good or bad for policing and if it should be opposed. And I guess I wanted to talk about how those arguments should be approached. 1/
It seems to me there are four ways to argue about this:
1. Based on preferences of the debating parties. 2. Based on (empirical) historical analysis of police capabilities and expectations of privacy. 3. Based on verifiable police need. 4. Based on democratic preferences. 2/
It seems like most public policy debate has focused on (1). I don’t think this is worth very much.
I’m interested in (2) and think we need more data, but I’m pretty confident it will tell a story of (exponentially) increasing police capability and individual loss of privacy. 3/
This letter is pretty amazing. It’s from a Senate Armed Services member explaining how they’re going to build the infrastructure to monitor most Internet users, network-wide using private DNS metadata.
For people who don’t know what DNS (Domain Name System) is: it’s basically the telephone directory for the Internet. Anytime one computer talks to another, it asks a DNS resolver to look up the other computer’s Internet address. These lookup requests are “metadata”…
… such that if you can buy the records of these lookups, you have effectively a “god’s eye view” over the entire Internet. Naturally the private companies who collect this data are happy to sell it off like it’s frozen orange juice.
Why Apple’s announcements today are a big deal, a thread. 1/
First: Apple has spent years building the infrastructure to build end-to-end backup for iCloud. This means backup where only you, and not Apple, hackers or the government, can access your own data. 2/
However: despite deploying the infrastructure to do this as far back as 2016, Apple limited the set of end-to-end encrypted data to things like passwords and your web history. Your text messages, photos, notes etc. were all accessible to someone who could get into iCloud. 3/
Looks like Apple is rolling out opt-in end to end encryption for iCloud backups. apple.com/newsroom/2022/…
This will cover every kind of data except for iCloud Mail, Contacts, Calendars: features that require server access.
I spoke with Apple earlier this morning about this proposal, and I was pretty impressed by what they’ve done. Unfortunately I’m about to have a dentist look in my mouth so it will have to wait.
There was a time before you all became obsessed with ChatGPT and there will be a time afterwards.
I know this is going to be one of those “640KB is enough” tweets, but we’ve had GPT models for years and DALL-E type models for a year. What are the real-world impacts we’re seeing (or will soon see) from this tech? Slightly more spam?
I think it’s pretty cool that we can erase unsightly objects from photos using ML, and GitHub Copilot seems like it might take some time out of programmers’ days. Maybe?