How good is Fediverse security?
2 Boosts
0 Likes
@evan@cosocial.ca I'd say the answer is multi-faceted.
I answered Fair.
The reasons why:
- LLMs are making it almost trivial to find security vulnerabilities. When they are easy to discover, they are only marginally harder to exploit.
- There exists an underlying current of AI resistance on the open social web. This of course, varies by subculture, but could affect many fediverse projects, who may be hesitant to allow AI-assisted contributions, or ban them outright.
- This indirectly puts them at risk of exploitation for merely standing on principle against AI.
We (@nodebb@fosstodon.org) have spent the better part of the past 6 months just patching holes in a product we thought was solid. It was humbling as hell.
@evan Define security.
@evan On here you get to have your posts scraped by thousands of bots, rather than just one big corporation.
@evan I think for it to be good, a lot more of your data needs to be protected from your host, and the hosts of instances you interact with, as standard. I think it's possible to get there though.
@evan You can’t really generalise like that. The Fediverse consists of so many different platforms – that is, pieces of software – that you’d have to be able to say something about each one individually. But you know that better than I do.
@hiker I disagree that it's impossible to reason about complex systems, with lots of connections and heterogeneous parts, and even to make statements about them. People do this all the time.
@evan each single instance may only be so-so in terms of security best practices but a huge benefit of being distributed is that any single security fuckup on an instance only affects a small number of people. the most effective way to prevent large amounts of data from being stolen is to not have a lot of data stored in a single place. So overall, pretty good.
@evan
in what context?
encrypted-at-rest storage? not usually
private messages that admins can't read? nope
end-to-end encryption? nah
assurance that messages are from who they say they are from? perhaps
secure that pathological JSON payloads won't screw up arbitrary clients? unlikely
secure against (D)DOS attacks? unlikely
is the Mastodon-proper™ sprawl of codebase well-audited? I have my doubts
GoToSocial, Snac2, etc? They tend to be smaller with fewer moving parts, so I have a bit more faith/trust in their integrity.