Mastodon, an alternative social network to Twitter, has a serious problem with child sexual abuse material according to researchers from Stanford University. In just two days, researchers found over 100 instances of known CSAM across over 325,000 posts on Mastodon. The researchers found hundreds of posts containing CSAM related hashtags and links pointing to CSAM trading and grooming of minors. One Mastodon server was even taken down for a period of time due to CSAM being posted. The researchers suggest that decentralized networks like Mastodon need to implement more robust moderation tools and reporting mechanisms to address the prevalence of CSAM.
Hi, since Mastodon is no longer acceptable due to the 0.04 percent of instances found to have abusive material, would someone please suggest the alternative social network with 0 percent of these incidents? Companies like Facebook and Twitter are driven by shareholders and greed, Mastodon is a community effort and you’ll certainly find bad actors there, but I feel less dirty contributing to a community project, versus helping billionaires like Zuck and Elon line their pockets harvesting my data.
“massive child abuse material problem”
“112 instances of known CSAM across 325,000 posts”
While any instance is unacceptable, does 112/325,000 constitute a “massive problem”?
0.0000034462% of posts are unacceptable! Massive problem!
You moved the period in the wrong direction. It’s
0.034462%.That’s just the material they knew was CSAM from previous investigations.
There were also 713 uses of the top 20 CSAM-related hashtags across the Fediverse on posts that contained media, as well as 1,217 text-only posts that pointed to “off-site CSAM trading or grooming of minors.” The study notes that the open posting of CSAM is “disturbingly prevalent.”
While the study itself is a good read and I agree with the conclusions—Mastodon, and decentralized social media need better moderation tools—it’s hard to not read the Verge headline as misleading. One of the study authors gives more context here https://hachyderm.io/@det/110769470058276368. Basically most of the hits came from a large Japanese instance that no one federates with; the author even calls out that the blunt instrument most Mastodon admins use is to blanket defederate with instances hosted in Japan due to their more lax (than the US) laws around CSAM. But the headline seems to imply that there’s a giant seedy underbelly to places like mastodon.social[1] that are rife with abuse material. I suppose that’s a marketing problem of federated software in general.
- There is a seedy underbelly of mainstream Mastodon instances, but it’s mostly people telling you how you’re supposed to use Mastodon if you previously used Twitter.
The person outright rejects defederation as a solution when it IS the solution, if an instance is in favor of this kind of thing you don’t want to federate with them, period.
I also find worrying the amount of calls for a “Fediverse police” in that thread, scanning every image that gets uploaded to your instance with a 3rd party tool is an issue too, on one side you definitely don’t want this kinda shit to even touch your servers and on the other you don’t want anybody dictating that, say, anti-union or similar memes are marked, denounced and the person who made them marked, targeted and receiving a nice Pinkerton visit.
This is a complicated problem.
Edit: I see somebody suggested checking the observations against the common and well used Mastodon blocklists, to see if the shit is contained on defederated instances, and the author said this was something they wanted to check, so i hope there’s a followup
The person outright rejects defederation as a solution when it IS the solution
It’s the solution in the sense that it removes it from view of users of the mainstream instances. It is not a solution to the overall problem of CSAM and the child abuse that creates such material. There is an argument to be made that is the only responsibility of instance admins, and that past that is the responsibility of law enforcement. This is sensible, but it invites law enforcement to start overtly trawling the Fediverse for offending content, and create an uncomfortable situation for admins and users, as they will go after admins who simply do not have the tools to effectively monitor for CSAM.
Defederation also obviously does not prevent users of the instance from posting CSAM. Admins even unknowingly having CSAM on their instance can easily lead to the admins being prosecuted and the instance taken down. Section 230 does not apply to material illegal on a federal level, and SESTA requires removal of material that violates even state level sex trafficking laws.
In my opinion the biggest issue the author points out is that cached materials are sometimes retained even after moderator action. Which honestly just sounds like a straight up bug more than anything. Though if I were running an instance, the feds showing up at my door with a warrant because I’ve been accidentally distributing CSAM would be my nightmare scenario. And of course jurisdiction plays a part, too: an American user on a Canadian server might see drawn depictions of sexualized minors, think “weird but not illegal,” and now the Canadian admin has content that’s illegal in Canada on their Canadian server and has no idea.
IMO I think the best solution to this is something similar to what Renaud Chaput (Mastodon’s resident infra boffin) described in his recent blog post. Effectively, give admins a way to hand this off to pluggable third-party services. Admins that are worried about this sort of thing can then have some degree of safety via e.g. PhotoDNA, whereas others can take on additional risk and preserve additional privacy.
All that said: yeah the headline makes it sound like .social is some 8chan-esque hellhole, whereas in reality my feed is 99% German programmers sharing milquetoast political takes.
Effectively, give admins a way to hand this off to pluggable third-party services.
https://blog.cloudflare.com/the-csam-scanning-tool/
https://developers.cloudflare.com/cache/reference/csam-scanning/
(From just a quick survey, the smaller open source CSAM scanning projects seem to mostly have stalled, and the government hosted ones are in perpetual design and planning).
Spitballing here, but having a corresponding images.lemmy.tld stood up along side each Lemmy instance where one could put that site behind cloud flare could be a workable approach.
While I agree that CSAM material needs to be addressed it’s also worth pointing out that cloudflare has some privacy issues.
Pedos that got banned from platforms turn to other platform who hasnt done it yet
In other news: the sky is blue
While white knights propose ways to control everyone everywhere everytime, in the name of catching the pedos who will just hop to the next platform (or have already).
I for one am all for instances being forcibly taken down by police if they can’t moderate CSAM appropriately.
Moderation is a very real challenge. The internet at large aimed to solved it by centralizing everything to a few mega corps with AI moderation. The fediverse aims to solve it by keeping instances small and holding both mods and users accountable.
I’m not fully sure about the logic and perhaps hinted conclusions here. The internet itself is a network with major CSAM problems (so maybe we shouldn’t use it?).
The internet itself is a network with major CSAM problems
Is it, though?
Over the last year, I’ve seen several reports on TV of IRL group abuse of children, by other children… which left everyone scratching their heads as to what to do since none of the perpetrators are legally imputable.
During that same time, I’ve seen exactly 0 (zero) instances of CSAM on the Internet.
Sounds to me like IRL has a major CSAM, and general sex abuse, problem.
It doesn’t help to bring whataboutism into this discussion. This is a known problem with the open nature of federation. So is bigotry and hate speech. To address these problems, it’s important to first acknowledge that they exist.
Also, since fed is still in the early stages, now is the time to experiment with mechanisms to control them. Saying that the problem is innate to networks is only sweeping it under the rug. At some point there will be a watershed event that’ll force these conversations anyway.
The challenge is in moderating such content without being ham-fisted. I must admit I have absolutely no idea how, this is just my read of the situation.
@mudeth @pglpm you really don’t beyond our current tools and reporting to authorities.
This is not a single monolithic platform, it’s like attributing the bad behavior of some websites to HTTP.
Our existing moderation tools are already remarkably robust and defederating is absolutely how this is approached. If a server shares content that’s illegal in your country (or otherwise just objectionable) and they have no interest in self-moderating, you stop federating with them.
Moderation is not about stamping out the existence of these things, it’s about protecting your users from them.
If they’re not willing to take action against this material on their servers, then the only thing further that can be done is reporting it to the authorities or the court of public opinion.
Maybe my comment wasn’t clear or you misread it. It wasn’t meant to be sarcastic. Obviously there’s a problem and we want (not just need) to do something about it. But it’s also important to be careful about how the problem is presented - and manipulated - and about how fingers are pointed. One can’t point a finger at “Mastodon” the same way one could point it at “Twitter”. Doing so has some similarities to pointing a finger at the http protocol.
Edit: see for instance the comment by @[email protected] to this post.
Understood, thanks. Yes I did misread it as sarcasm. Thanks for clearing that up :)
However I disagree with @[email protected] in that Lemmy, and the Fediverse, are interfaced with as monolithic entities. Not just by people from the outside, but even by its own users. There are people here saying how they love the community on Lemmy for example. It’s just the way people group things, and no amount of technical explanation will prevent this semantic grouping.
For example, the person who was arrested for CSAM recently was running a Tor exit node, but that didn’t help his case. As shiri pointed out, defederation works for black-and-white cases. But what about in cases like disagreement, where things are a bit more gray? Like hard political viewpoints? We’ve already seen the open internet devolve into bubbles with no productive discourse. Federation has a unique opportunity to solve that problem starting from scratch, and learning from previous mistakes. Defed is not the solution, it isn’t granular enough for one.
Another problem defederation is that it is after-the-fact and depends on moderators and admins. There will inevitably be a backlog (pointed out in the article). With enough community reports, could there be a holding-cell style mechanism in federated networks? I think there is space to explore this deeper, and the study does the useful job of pointing out liabilities in the current state-of-the-art.
Another way to look at it is: How would you solve this problem with email?
The reality is, there is no way to solve the problem of moderation across disparate servers without some unified point of contact. With any form of federation, your options are:
- close-source the protocol, api, and implementation and have the creator be the final arbiter, either by proxy of code, or by having a back door
- Have every instance agree to a singular set of rules/admins
- Don’t and just let the instances decide where to draw lines.
The reality is, any federated system is gonna have these issues, and as long as the protocol is open, anyone can implement any instance on top of it they want. It would be wonderful to solve this issue “properly”, but it’s like dealing with encryption. You can’t force bad people to play by the rules, and any attempt to do so breaks the fundamental purpose of these systems.
This is exactly what I thought. The story here is that the human race has a massive child abuse material problem.
The problem is even bigger: some places (ejem Reddit) you will get deplatformed for explaining and documenting why there is a problem.
(even here, I’ll censor myself, and hopefully restrict to content not too hard to moderate)
Does this come to the surprise of anyone? This is the case on large social networks and decentralised networks by definition are free and open for anyone to moderate how they please. Unfortunately including pedos
So instances that are actually supporting CSAM material can and should be dealt with by law enforcement. That much is simple (and I’m surprised it hasn’t been done with certain … instances, to be honest). But I think the apparently less clearly solved issues have known and working solutions that apply to other parts of the web as well. No content moderation is perfect, but in general, if admins are acting in good faith, I don’t think there should be too much of a problem:
- For when federation inadvertently spreads some of the material through to other instances’ databases: Isn’t this the same situation as when ISP’s used to cache web traffic to save on bandwidth costs? In that situation, too, browsed web pages would end up in the ISP’s cache which could then harbour whatever material the user was looking at. As I recall, the ISP would just ban CSAM and other illegal material in their terms of service, and remove anyone reported as violating the rule, and that sufficed.
- As for “bad” instances/users: It’s impossible to block all instances and all users that might disseminate this material as you’d have to go to a “block everything, then allow known entities” rule which would break the Fediverse model. Again, users or site admins found to be acting in bad faith should be blocked and reported (either automatically or manually). Some may slip through the net, but as long as admins are seen to be doing the best they can, that should be enough.
There seem to be concerns about “surveillance” of material on Mastodon, which strikes me as a bit odd. Mastodon isn’t a private platform. People who want private messaging should use an E2EE messaging app like Signal, not a social networking platform like Mastodon (or Twitter, Threads etc.). Mastodon data is already public and is likely already being surveilled, and will be so regardless of what anyone involved with the network wants, because there’s no access control on it anyway. Having Mastodon itself contain code to keep the network clean, even if it only applies to part of the network, just allows those Mastodon admins who are running that part of the code to take some of the responsibility on themselves for doing so, reducing the temptation for third parties to do it for them.
Yeah I recall that the Japanese instances have a big problem with that shit. As for the rest of us, Facebook actually open sourced some efficient hashing algorithms for use for dealing with CSAM; Fediverse platforms could implement these, which would just leave the issue of getting an image hash database to check against. All the big platforms could probably chip in to get access to one of those private databases and then release a public service for use with the ecosystem.
Facebook actually being good for once? Unheard of
As much as we can (and should) lambast Facebook/Meta’s C-Suite for terrible decisions, their engineers are generally pretty legit.
They actually contribute a lot of useful stuff to the web dev world, like React.js. It’s just all the other shit they do that’s awful.
That’d be useless though, because first, it’d probably opt-in via configuration settings and even if it wasn’t, people would just fork and modify the code base or simply switch to another ActivityPub implementation.
We’re not gonna fix society using tech unless we’re all hooked up to some all knowing AI under government control.
That’s not the point. Yes, child porn sites can host child porn. Other sites/instances can’t stop that. But what other instances can stop, is redistributing said child porn. And for that purpose, such technology would be useful.
researchers found 112 instances of known CSAM across 325,000 posts on the platform
So you’re willing to vacuum up the hashes of every image file uploaded on thousands of decentralized systems into a centralized systems (that is out of “our” control and coupled with direct access for law enforcement and corporations) to prevent the distribution of 0.034% of files that are CSAM and that could just as well be reported and deleted by admins and moderators? Remember how Snowden warned us about metadata?
If you think that’s a wise tradeoff, I guess, go ahead. But then I’d have to question the entire goal of being decentralized in the first place. If it’s all about “a billionare can’t wreak havok upon my social network”, then yeah, I guess decentralization helps a bit but even that remains to be seen.
But if you’re actually willing to do that, you’d probably also be in favor of having government backdoors into chat encryption (and thus rendering the entire concept moot, because you can’t have backdoors that cannot be discovered by other nefarious actors) and even more censorship-resistant systems like Tor because evil people use it to exchange CSAM anonymously as well?
I don’t know how you get the impression that this increases censorship.
Instance admin already manually block content. And they are already able to do that to any extend they wish to do.
This tool would simply automate that process.
Admins would not gain or lose any ability to block content. Identifying child porn would simply be easier.
(Imagine an admin going to their database and doing a CTRL+F with the term “child porn”, and then going through the posts to find offending ones. But instead of CTRL+F it’s an AI.)
(For some reason I don’t get a notification when you answer my comment. Is that a known issue? Did you block me or something?)
I’m referring to the CSAM scanning systems that are outside of the control of almost anyone except governments, three letter agencies, other law enforcement and parts of the private sector.
These systems must be fed every hash of every file submitted to as many instances as possible to be efficient with close to no oversight or public scrutiny.
Pass.
Edit: I’m not blocking you but I noticed intermittent connectivity issues on lemmy.ml today, possibly around the time where I replied.
I don’t know how you get the impression that this increases censorship.
This tool would simply automate that process.
Well… precisely?
Censorship is any removal of material considered “undesirable”, whether you agree with why it is considered “undesirable” or not.
If you want more censorship of “material that you personally consider undesirable”, then just say so, don’t hide behind some disingenuous “but it isn’t censorship”. Then we can discuss the merits of that classification, and of the means proposed to achieve such censorship.
You seem to be missing my point.
This tool would not increase censorship.
Admins are already able to implement all censorship they want.
Admins are already able to block left-wing opinions, right-wing opinions, child porn, normal porn.
And that already happens.
Lots of instances (like feddit.de) block pornographic content.
Lots of instances (like lemmy.blahaj.zone) block right-wing content.
It is already possible, and it is already happening.
An AI which can detect CSAM (and potentially other content) won’t change that. It will simply make the admins’ job easier.
I think you’re missing the opposite point.
An AI trained on a given instance’s admin decisions, would increase the same censorship the admins already apply. We can agree on that.
An AI trained by a third-party on unknown data (and actually illegal to be known) which can detect “CSAM (and potentially other content)”, would increase censorship of both CSAM… and of “potentially other content” out of the control, preferences or knowledge of the instance admins.
Using an external service to submit ALL content for an AI trained by a third-party to make a decision, not only allows the external service to collect ALL the content (not just the censored one), but also to change the decision parameters without previous notice, or any kind of oversight, and apply it to ALL content.
The problem is a difference between:
- instance modlog -> instance content filtered by instance AI -> makes similar decisions as instance admins
- [illegal to know dataset] -> third-party captures all content, feeds to undisclosed AI -> makes unknown decisions in the name of removing CSAM
One is an AI that can make mistakes, but mostly follows whatever an admin would do. The other, is a 100% surveillance state nightmare in the name of filtering 0.03% of content.
If you read the article, there’s actually more. The problem also isn’t just that they post the material directly onto Mastodon, they also use the platform to network.
More… or less, given that US-centric CSAM detectors mark AI, CG and drawings at the same level as IRL images.
Preventing the networking, is called defederation, that’s already there.
The researchers suggest that decentralized networks like Mastodon need to implement more robust moderation tools and reporting mechanisms to address the prevalence of CSAM.
I agree, but who’s going to pay for it? Those aren’t just freely available additions to any application that you only need to toggle on.
The researchers can’t be taken seriously if they don’t acknowledge that you can’t force free software to do something you don’t want it to.
Even if we started way down at the stack and we added a CSAM hash scanner to the Linux kernel, people would just fork the kernel and use their own build without it.
Same goes for nginx or any other web server or web proxy. Same goes for Tor. Same goes for Mastodon or any other Fedi/ActivityPub implementation.
It. Does. Not*. Work.
* Please, prove me wrong, I’m not all knowing, but short of total surveillance, I see no technical solution to this.
I agree, but who’s going to pay for it?
How about police/the tax payer?
If university researchers can find the stuff, then police can find it too. There should be an established way to flag the user (or even the entire instance) so that content can be removed from the fediverse while simultaneously asking for all data that is available to try to catch the criminals.
And of course, if regular users come across anything illegal they will report it too, and it should be removed quickly (I’d hope immediately in many cases, especially if the post was by a brand new/untrusted account).
A decentralised platform like the Fediverses won’t easily work with nation states and their taxes. Even with Wikipedia today, it’s not funded directly via any government - but rather by certain universities giving some money to it + all the private doners.
And even if we get that working, power politics will mess this up like so often when things actually get troublesome.
It might be interesting to explore cryptocurrencies as for donations here though. They do have international liquidity and they can’t be misused foe power politics.
One of the problems with the fediverse is that each server keeps its own copy of the content. It is definitely a worry that bad actors push content to federated servers to get them taken down due to the content they now are storing.
What’s the reason for that? Caching purposes?
General idea is that if there is only one copy, taking something down is knocking that server out of service.
I think so. My Lemmy instance for example is currently storing several gigabytes of images in my cloud buckets, but with my 4 users I’m reasonably confident it didn’t all come from us.
This is why I disabled that feature on my Lemmy instance.
