Authorized Fetch Circumvented by Alt-Right Developers
Authorized Fetch Circumvented by Alt-Right Developers
A controversial developer circumvented one of Mastodon's primary tools for blocking bad actors, all so that his servers could connect to Threads.
Authorized Fetch (also referred to as Secure Mode in Mastodon) was recently circumvented by a stupidly easy solution: just sign your fetch requests with some other domain name.
Repeat after me: anything I write on the internet should be treated as public information. If I want to keep any conversation private, I will not post it in a public website.
I agree with you, however there are issues with not just privacy but also authenticity. I should be able to post as me, even in public, and have a way to prove it. Nobody else should be posting information as me, if that makes sense.
For that, we should start bringing our own private keys to the server, instead of trusting the server to control everything.
And if we start doing that, pretty soon we will end up asking ourselves why do we need the server in the first place, and we will evolve to something like what nostr is doing.
I'm all for it.
Seriously. Bobthenazi could just go to an aligned server and make an account Bobthenotzi and boom -- perfectly able to follow whoever he wants.
To add a bit of important nuance to this idea (particularly how this argument comes up with regards to threads). This does not apply to legal rights over your content. That is to say, of course you should treat any information you put out there as out of your control with regards to access but if somebody tries to claim legal rights over your content they are probably breaking the law.
anything I write on the internet should be treated as my private information. If I want to keep any conversation private, I will still post it in a public website.
EDIT: I'm so sorry that my stupid comment offended some people. Always forget how special some people can be on this website. Once again I'm sorry for my lack of better judgement.
I don't think your comment was offensive per se. It was just ridiculously naive. If we are trying to build practical tools, they have to fit how things work in the real world, not how they work in anybody's dreams. If you want to have private conversations on a public website, use encryption.
And so, once again, people discover the unsolvable dilemma of DRM.
You can't both publish your data where it can be seen by computers that are not under your control and somehow keep control of that data. Anything that purports to do so is either a temporary bandaid soon to be bypassed or nothing but placebo to begin with.
What I think a lot of conversations about privacy and security on the Fediverse miss is that the Fediverse is radically public.
A protocol that sends everything you share to a long list of servers that haven't been pre-screened and could be anything from a professionally-managed instance of vanilla Mastodon to an ad hoc, informally-specified, bug-ridden, slow implementation of half of ActivityPub running on a jailbroken smart light bulb can only ever be radically public. It's possible to block most interactions with someone you don't want to talk to, but not to reliably prevent them from seeing content you share to anything more than a short list of vetted followers.
There probably isn't any reasonable way to change that while keeping the open federation model, though it's possible to build closed networks on top of ActivityPub for those who want the formats it supports for a curated group. This isn't a problem to be solved in my view, but an inherent reality: the Fediverse is for things you want to make public.
Exactly! The only way that we can make sure that the Internet is not controlled by anyone is to make it available for everyone. If we are fighting for an open internet, we need to understand that this type of thing will be part of the package.
We, by which I mean some loose group of people who want decentralized tools to thrive should also be building things for secure, private communication, and we are. Matrix, for example offers strongly end-to-end encrypted federated chat rooms and private messages. It also has a kind of rough UX and, IIRC resource-intensive server software. We should work toward improving that.
I'm not advocating against privacy at all. I want people to understand as clearly as possible that Mastodon, Lemmy, and anything that works like them isn't private and can't be private when part of an open federated network so they can decide whether that's a good fit for how they're using it. The block evasion described in the link is just run a server on a domain that isn't blocked, and I imagine any other mitigations bolted onto Mastodon that don't break open federation will be little better.
I'm kind of tired of social networks offering even the pretense of privacy. Just loudly proclaim that everything is public but clients can filter out shit you don't wanna see.
That doesn't work for vulnerable minorities. Manually filtering each shitty person after you step in their shit gets old. Coupled with the fact that not shutting down shitty people just means more shitty people are likely to turn up.
It's not sustainable
I think in this context it's meant on a technical level: as far as the fediverse is concerned, there's not a whole lot instances can do. Anyone can just spin up an instance and bypass blocks unless it works on an allowlist basis, which is kind of incompatible with the fediverse if we really want to achieve a reasonable amount of decentralization.
I agree that we shouldn't pretend it's safe for minorities: it's not. If you're a minority joining Mastodon or Lemmy or Mbin, you need to be aware that blocking people and instances has limitations. You can't make your profile entirely private like one would do on Twitter or any of Meta's products. It's all public.
You can hide the bad people from the users but you can't really hide the users from the bad people. You can't even stop people from replying to you on another instance. You can refuse to accept the message on the user's instance, but the other instance can still add comments that don't federate out. Which is kind of worse because it can lead to side discussions you have no way of seeing or participate in to defend yourself and they can be saying a lot of awful things.
It's the unfortunate reality. Social networks simply cannot in offer privacy. If they were upfront about it, then people could make rational decisions about what they share.
But instead they (including Mastodon) pretend like they can offer privacy, when they in fact cannot, resulting in people sharing things that they would not otherwise share.
It’s not sustainable to keep offering poorly designed solutions. People need to understand some basic things about the system they're using. The fediverse isn't a private space and fediverse developers shouldn't be advertising pseudo-private features as private or secure.
I have no idea what any of that meant.
I still could use an ELI5 about what this authorized fetch feature was supposed to do. Was it supposed to basically disengage the Mastodon network from Threads? To stop Threads crap from showing up on Mastodon? Or to stop Mastodon discussions from showing up in Threads? Or something different?
Authorised Fetch existed long before Instagram Threads. When it is turned on, an instance will require any other server to sign their request to fetch any post. This prevents "leaking" of posts via ActivityPub to blocked instances.
This setting is turned off by default, because some software are incompatible with it (like /kbin, Pixelfed before June 2023, maybe Lemmy too), because it makes server load higher, and it may make some replies missing (at least on microblogging side).
When it is turned on, an instance will require any other server to sign their request to fetch any post. This prevents “leaking” of posts via ActivityPub to blocked instances.
Oh I see. Yeah that sounds pretty hopeless. Does it use the fetching site's domain validated TLS certificate? Is the idea to permit fetching unless the fetching domain is on a blacklist? If yes, someone didn't have their thinking cap on. The whole concept is dumb though, there is no way to prevent posts from leaking. The saying is that once 3 people know a secret, it is no longer secret.
Stop asking for pseuso-privacy features. The Fediverse is public by nature. Any "measures" to control access to the public posts on it are just lying to users.
Server owners should be able to control who can access their servers - but that is NOT - and should NOT be - treated as a privacy feature.
“The Net interprets censorship as damage and routes around it.” - John Gilmore