1. 33
  1. 8

    Nothing hugely shocking here. If you have a decentralized system without end to end crypto then servers can read all your stuff, its the same with email and gmail scanning all of your emails.

    1. 6

      Which is why we shouldn’t build decentralised (or centralised) systems without end-to-end crypto any longer.

      There’s no reason why something like Mastodon couldn’t have anonymous (unsigned, unencrypted), public (signed, unencrypted), group (signed, encrypted to a group — ‘friends’ is merely one group), and unlisted (signed, encrypted) posts. Yes, there are some key management challenges (particularly around key management & re-encryption as one adds & deletes friends), but they are no insurmountable.

      I strongly believe that writing systems without cryptographically-strong privacy in 2018 is an error.

      1. 8

        Secure Scuttlebutt is a pretty good example of this, you have public messages and private messages. If a message is private then it is encrypted and only people mentioned in the post can decrypt the message. But ssb does have serious key management issues.

        1. 5

          What are the key management issues? I was just coming here to mention ssb, but I’m very new to it and was unaware of this. Can you share more?

          1. 4

            Well off the top of my hat, key management issues arise whenever you try to use it across multiple machines. Now you could manually copy the key from machine to machine, but if you ever use two machines simultaneously it creates a sort of fork in your identity on the network, which causes plenty of trouble.

            There are a few solutions under research, most notable a master / slave system, but last time I checked it was still very much in the design phase.

        2. 4

          This is easily said, but both end to end crypto and key management add a huge amount of complexity to the system. If you need the privacy that e2e can provide, this is of course worth it, but it’s not at all clear that every service needs this. The fediverse is meant for public and targeted messages, not private ones. For those usecases, people can easily use e2e encrypted systems like matrix or gpg.

          1. 3

            Hear hear! I think everyone has this vision of a perfect crypt-opia where we can conduct our social networking safe from the prying eyes of government or BigCorps, but the realities of making this happen are as you say not at all trivial.

            It’s a great goal, and one I think people should continue working towards, but the logistics are hard.

            1. 3

              Privacy and social media are kind of at odds with each other anyway. People want to share their posts with the world but also not have that data used against them. If you didn’t want everyone to know then you shouldn’t be sharing it.

              1. 1

                I don’t know if I agree. When I publish toots on Mastodon, all they know is that feoh@amicable.feoh.org said blah blah blah.

                When I use Facebook, they are collecting a SUPER rich trove of demographic data on me, cross referencing it with other commercial sources (my employer for one :) and linking it in with my “social graph” where my friends data is taken into account. It’s the difference between a linked list of nodes with 2 or 3 fields and a full on acyclic graph with zillions of nodes and zillions more connections.

                1. 1

                  all they know is that feoh@amicable.feoh.org said blah blah blah.

                  Anyone can also see who you are following, who you reply too, whos posts you like, what kind of content you like and then draw a graph based on this data. The main thing you lose is the tracking using apps to see more than what you post but a huge huge amount of data anyone can see can be used to track you and build a profile on you.

                  1. 1

                    By ‘anyone’ you mean ‘any Fediverse user’ right? Also there’s a huge difference between having to scrap the correlate vast gobs of data yourself and having it handed to you for analysis on a silver platter by the platform.

                    Anyway, this is silly. I agree that social media is at odds with privacy to an extent, but some platforms are factually, provably better than others.

            2. 1

              also there’s nothing stopping you from using clientside tools that provide this.

            3. 3

              I totally disagree. I think there is a place in the world for social network protected by crypto, and also for those that aren’t.

              Let’s not let the perfect be the enemy of the good.

              1. 1

                How would you do this while still allowing mastodon to be used from a web interface? If it’s implemented using javascript you’re in the exact same situation of having to trust the instance administrator.

                1. 1

                  How would you do this while still allowing mastodon to be used from a web interface?

                  I’d either use a native client, or a web client running on localhost. It’s the only way to assure privacy & security.

                  1. 0

                    Exactly. You said that there’s no reason why it couldn’t do this, there’s your reason.

              2. 3

                agreed. It seems mostly useful for like novices.

                1. 1

                  What scares me is the resignation to this state of things.

                  1. 4

                    If you store things on other people’s servers they are on other people’s servers. I don’t see how this statement is a resignation. If you want your posts to be private in the fediverse, encrypt it. If you want your emails, posts,etc to be private, encrypt them.

                    1. 5

                      I was not talking about @mercer article: as you said it can be pretty useful for novices.

                      What scares me is that we could design something better, but there is not much research about the topic.

                      No one really try to challenge the status quo with original engineering solutions, in a sort of resignation.

                      At best, people are waiting for mathematicians to create a cheap fully homomorphic encryption scheme.

                      But I’m afraid it’s not lazyness, but lack of vision, interest and hope.

                      1. 5

                        Vision, interest, and hope are not valid inputs to compilers.

                        I think a reasonable compromise in new system design (taken in some side projects of mine) is to assume that the channels of communication are compromised by hostile actors, that storage exists in the datacenters of hostile actors who are actively trying to munge through the contents, and that mere possession of encrypted material is of significant interest to the hostile actors.

                        You end up with a sort of “I am Spartacus” setup for communication systems under those constraints, where everybody by definition has open-access to all communications but all communications are also encrypted such that if you have a key you can read it and otherwise you are just providing storage–and because everybody has copies of the content, the metadata of how it moves through the system is not super interesting. Of course, the flipside is that participation in such a system is almost always a red flag.

                        1. 1

                          Well… vision alone gave UNIX pipelines. And stacks. And timesharing systems… ;-)
                          Interest gave us Linux. And hope gave us GNU.

                          But, your system description look interesting… can you share links to some free software designed that way?

                        2. 3

                          If you can’t read the code on the server, and you can’t, then you can’t know it was actually encrypted. The only thing you can do is end to end encryption, which you can already do on top of all of these existing services. What we need is education of the tools that already exist and also improving ease of use. The moment you put the tech on the server you’ve already lost. Otherwise the tech you’re describing already exists.

                          1. 3

                            I agree with you about education. I deeply agree.

                            But with fully homomorphic encryption you can know it’s encrypted even without seeing the code.

                            I’m not entirely sure that no other mitigation is possible: my insight is that too few have tried to challenge the http/dns/browser/javascript stack to get a chance to find a solution.

                            My bet is that we just need to open our minds.

                            Still, you are right: there’s no cloud, just another person’s computer… ;-)

                2. 4

                  Good article. I really like how even handed it is. Sometimes folks who write bits like this tend to be super polarized in their opinion and represent things as black and white, whereas the world is almost always painted in shades of gray.

                  1. 2

                    Plus the author is hilarious. Pretty much the whole article needs to be wrapped in a sarcasm tag.