5 ms·

You laid out a number of points there, and I think some of them are indicative of what's really holding back security for the masses. "Matching a name with a s

You laid out a number of points there, and I think some of them are indicative of what's really holding back security for the masses. "Matching a name with a social media is the wrong way to lookup others" -- this is the one I take greatest issue with, because it's the way we know people now. All the people I collaborate with online now, I know them by those exact names. Coworkers, remote collaborators, friends on instagram, famous software engineers (whose work I might want to consume), even my own brothers... I know them by a set of these identities. This is how I know people. To illustrate this point: if my own brother wrote me a message and posted it simultaneously on Twitter, Facebook, and LinkedIn, I would think "yeah, that's my brother." Moreover, and more importantly, if he left those posts up publicly, over time the strength of my conviction that he actually wrote the messages (and it wasn't a short-term compromise) would grow. This works even for people you've maybe never met in real life. If hacker news user "jashkenas" announces on HN that he has a new version of CoffeeScript, I expect that's him. If he pairs it with a tweet, then I really expect it's him. The fact that this method works for possible strangers and loved ones is very powerful. Remember, this isn't just about secure messaging. If someone doesn't have any such identities, then there's always still the fallback: exchange an identifier in person. The alternatives are so nasty -- in person meeting, coordinating enterprisey apps, and so on -- we'll never get anywhere with this kind of thing. The solution I think you are suggesting requires a lot of human effort to figure out if you have the person you want. This is one of the danger points of PKI. And one of the inconvenience points. "Leaving people to manage their own private keys is worse for security than having them managed by others." Part of me wonders if you're just trolling. I don't understand this, and I've read your paragraph a few times. There is the confusion argument - that people don't understand how to manage them, which is a problem we're tackling. Our argument is that it's possible to build something usable (finally!) where people do in fact own their own keys. But you go on to say "software glitches" and "updates" are dangerous for individuals who have device keys, and then suggest they'd be ok if their enterprises / social networks managed their private keys for them. In this case, you're talking about expanding the threats considerably, while still leaving client apps that need to be updated and do the work.

irq-1 11y ago
To use your example, your brother posts a message on Twitter and you don't know your brothers Twitter username. If you're confident that it's your brother posting, then there's no problem. If you're uncertain that it's your brother, Keybase will let you verify it's him -- but that's a problem. Imagine a political dissident living in a dictatorship and it's the government who uses Keybase to match a Twitter username to a Facebook account. We don't want accounts tied together. People need multiple identities online. How do you get a persons key, and how do you know it's the right key? I think the answer will work like email addresses do now: someone will tell you a little info (name + hash@org) and you'll lookup their key.[0] Then you'll verify it as appropriate to the situation. That kind of system would allow and encourage multiple identities, and doesn't encourage posting PII. To "bring public key crypto to everyone" you don't need to build this kind of directory (where you verify accounts.) You can instead get companies to participate in a PKI that they control (like they control email, DNS, http, etc..) and then they'll integrate it into existing software, provide it to users, experiment with UX and client designs, etc... Making an open source, server-side solution lets you turn 'Sign in with Facebook' into an ally that not only gets Facebook to provide keys and integrate crypto, but invest in servers and software. Last, many people share a sentiment that a private key shouldn't leave our computers or phones (or fobs) to be managed by others. It's a reasonable attitude, but I think getting "crypto to everyone" and widely implemented is more important, and a managed system today doesn't preclude unmanaged keys in the future. [0] To expand on this example: I'm irq-1 ff38a9@keys.news.ycombinator.com, and at work I'm Alice 23ff45@work. If you lookup hash@org and find a string that looks right (a username or real name) then you get their public key. You have some confidence it's the right person, and are free to verify it further with any other means.
- dllthomas 11y ago
  I am with you 110% that people need to be able to have multiple identities online that are not linkable. That does not mean they should not be able to link multiple accounts to a single identity. That is to say, "one identity per person" is wrong; "one identity per account" is also wrong. Both of these in the sense that they should be permitted if that is what the person desires but should not be expected or (especially) enforced.

irq-1 11y ago

To use your example, your brother posts a message on Twitter and you don't know your brothers Twitter username. If you're confident that it's your brother posting, then there's no problem. If you're uncertain that it's your brother, Keybase will let you verify it's him -- but that's a problem. Imagine a political dissident living in a dictatorship and it's the government who uses Keybase to match a Twitter username to a Facebook account. We don't want accounts tied together. People need multiple identities online. How do you get a persons key, and how do you know it's the right key? I think the answer will work like email addresses do now: someone will tell you a little info (name + hash@org) and you'll lookup their key.[0] Then you'll verify it as appropriate to the situation. That kind of system would allow and encourage multiple identities, and doesn't encourage posting PII. To "bring public key crypto to everyone" you don't need to build this kind of directory (where you verify accounts.) You can instead get companies to participate in a PKI that they control (like they control email, DNS, http, etc..) and then they'll integrate it into existing software, provide it to users, experiment with UX and client designs, etc... Making an open source, server-side solution lets you turn 'Sign in with Facebook' into an ally that not only gets Facebook to provide keys and integrate crypto, but invest in servers and software. Last, many people share a sentiment that a private key shouldn't leave our computers or phones (or fobs) to be managed by others. It's a reasonable attitude, but I think getting "crypto to everyone" and widely implemented is more important, and a managed system today doesn't preclude unmanaged keys in the future. [0] To expand on this example: I'm irq-1 ff38a9@keys.news.ycombinator.com, and at work I'm Alice 23ff45@work. If you lookup hash@org and find a string that looks right (a username or real name) then you get their public key. You have some confidence it's the right person, and are free to verify it further with any other means.

dllthomas 11y ago
I am with you 110% that people need to be able to have multiple identities online that are not linkable. That does not mean they should not be able to link multiple accounts to a single identity. That is to say, "one identity per person" is wrong; "one identity per account" is also wrong. Both of these in the sense that they should be permitted if that is what the person desires but should not be expected or (especially) enforced.