117

Child Safety on Federated Social Media (purl.stanford.edu)

submitted 1 year ago by FlyingSquid@lemmy.world to c/fediverse@lemmy.world

73 comments fedilink hide all child comments

Not the best news in this report. We need to find ways to do more.

top 50 comments

sorted by: hot top controversial new old

[-] shrugal@lemm.ee 43 points 1 year ago* (last edited 1 year ago)

Why do they just mention absolute numbers, instead of comparing them to similar platforms? All they said was that there is CSAM on the Fediverse, but that's also true for centralized services and the internet as a whole. The important question is whether there is more or less CSAM on the Fediverse, no?

This makes it look very unscientific to me. The Fediverse might have a CSAM problem, but you wouldn't know it from this study.

[-] WhoRoger@lemmy.world 9 points 1 year ago* (last edited 1 year ago)

Fediverse also makes it potentially easier to scan for this stuff. You can just connect a new server to the network and if the material is on federated servers, then you can find it (probably). While if it's some private forum or even the dark web, I assume it's a lot more difficult.

The other thing is, most regular servers defederate from suspicious stuff already. Like pretty much nobody federates with that one shota instance, and they only serve drawm stuff (AFAIK). So I don't know if you can even say servers like that are a part of the Fediverse in the first place.

[-] shrugal@lemm.ee 3 points 1 year ago

That's what I thought as well. If the authors of this "study" were able to simply scan for it on the Fediverse, then what's stopping law enforcement units from doing the same? They can literally get a message everytime someone posts something on a suspicious instance.

[-] MyFairJulia@lemmy.world 33 points 1 year ago

Why would someone downvote this post? We have a problem and it's in our best interest to fix that.

[-] chaogomu@kbin.social 47 points 1 year ago

The report (if you can still find a working link) said that the vast majority of material that they found was drawn and animated, and hosted on one Mastodon instance out of Japan, where that shit is still legal.

Every time that little bit of truth comes up, someone reposts the broken link to the study, screaming about how it's the entire Fediverse riddled with child porn.

[-] MyFairJulia@lemmy.world 16 points 1 year ago

So basically we had a bad apple that was probably already defederated by everyone else.

[-] Kes@lemmy.blahaj.zone 12 points 1 year ago

Moreso an apple with controversial but not strictly CSAM material based in a country where it's content is legal. Actually, not even an apple; Lemmy and the fediverse aren't an entity. It's an open standard for anyone to use; you don't see the Modern Language Association being blamed for plagiarized essays written in MLA format, or the WHATWG being blamed because illegal sites are written in HTML, so it's not a fair comparison to say that Lemmy/the fediverse are responsible for what people do with their open standard either

[-] ZILtoid1991@kbin.social 6 points 1 year ago

It's Pawoo, Pixiv's (formerly) own instance, which is infamous for this kind of content, and those are still "just drawings" (unless some artists are using illegal real-life references).

[-] dustyData@lemmy.world 3 points 1 year ago

They're using Generative AI to create photo realistic renditions now, and causing everyone who finds out about it to have a moral crisis.

[-] ZILtoid1991@kbin.social 5 points 1 year ago

Well, that's a very different and way more concerning thing...

[-] Derproid@lemm.ee 3 points 1 year ago

... I mean ... idk ... If the argument is that the drawn version doesn't harm kids and gives pedos an outlet, is a ai generated version any different?

[-] brain_pan@infosec.pub 6 points 1 year ago

imo, the dicey part of the matter is "what amount of the AI's dataset is made up of actual images of children"

[-] Derproid@lemm.ee 2 points 1 year ago

Shit that is a good point.

[-] etrotta@kbin.social 6 points 1 year ago

Here's a link to the report: https://stacks.stanford.edu/file/druid:vb515nd6874/20230724-fediverse-csam-report.pdf
It is from 2023-07-24, so there's a considerable chance it is not the one you were thinking about?

[-] chaogomu@kbin.social 14 points 1 year ago

Since the release of Stable Diffusion 1.5, there has been a steady increase in the
prevalence of Computer-Generated CSAM (CG-CSAM) in online forums, with
increasing levels of realism.17 This content is highly prevalent on the Fediverse,
primarily on servers within Japanese jurisdiction.18 While CSAM is illegal in
Japan, its laws exclude computer-generated content as well as manga and anime.

Nope, seems to be the one. They lump the entire Fediverse together, even though most of the shit they found was in Japan.

The report notes 112 non-Japanese items found, which is a problem, but not a world shaking issue. There may be issues with federation and deletion orders, which is also an issue, but not a massive world shaking one.

Really, what the report seems to be about is the fact that moderation is hard. Bad actors will work around any moderation you put in place, so it's a constant game of whack-a-mole. The report doesn't understand this basic fact and pretends that no one is doing any moderation, and then they add in Japan.

[-] dustyData@lemmy.world 7 points 1 year ago* (last edited 1 year ago)

I can't seem to find the source for the report about it right now, but there's literal child porn being posted to Instagram. We don't see this kind of alarmist reports about it because it is not something new, foreign and flashy for the general public. All internet platforms are susceptible to this kind of misuse. The question is what moderation tools and strategies are in place to deal with that. Then there's stuff like on TOR where CSAM was used as a basis to discredit the use of the whole technology then it turned out that the biggest repository was an FBI honey pot operation.

[-] chaogomu@kbin.social 10 points 1 year ago

Buried in this very report, they note that Instagram and Twitter have vastly more (self generated) child porn than the Fediverse. But that's deep into section 4, which is on page 8. No one is going to read that far into the report, they might get through the intro, which is all doom and gloom about decentralized content.

[-] Metal_Zealot@lemmy.world 3 points 1 year ago

It's 4Chan's "9000 Penises" all over again

[-] shrugal@lemm.ee 18 points 1 year ago* (last edited 1 year ago)

The study doesn't compare their findings to any other platform, so we can't really tell if those numbers are good or bad. They just state the absolute numbers, without really going into to much detail about their searching process. So no, you can't draw the conclusion that the Fediverse has a CSAM problem, at least not from this study.

Of course that makes you wonder why they bothered to publish such a lackluster and alarmistic study.

[-] bumblebrainbee@lemmy.ml 3 points 1 year ago* (last edited 1 year ago)

Pretty sure any quantity of CSAM that isnt zero is bad....

[-] shrugal@lemm.ee 11 points 1 year ago* (last edited 1 year ago)

Yes, but you don't need a study to know that there will be some CSAM on the Fediverse. This is about whether using the Fediverse could make things worse, and just stating absolute numbers won't answer that question. What it does do is make it sound like the Fediverse is a magnet for CSAM, when in reality the opposite could be true.

Or to put it differently: If these numbers turn out to be lower than what you get on similar platforms, then this could actually be a good sign, even though they still aren't zero.

load more comments (1 replies)

[-] Aesthesiaphilia@kbin.social 12 points 1 year ago

Because it's another "WON'T SOMEONE THINK OF THE CHILDREN" hysteria bait post.

They found 112 images of cp in the whole Fediverse. That's a very small number. We're doing pretty good.

[-] etrotta@kbin.social 5 points 1 year ago

It is not "in the whole fediverse", it is out of approximately 325,000 posts analyzed over a two day period.
And that is just for known images that matched the hash.

Quoting the entire paragraph:

Out of approximately 325,000 posts analyzed over a two day period, we detected
112 instances of known CSAM, as well as 554 instances of content identified as
sexually explicit with highest confidence by Google SafeSearch in posts that also
matched hashtags or keywords commonly used by child exploitation communities.
We also found 713 uses of the top 20 CSAM-related hashtags on the Fediverse
on posts containing media, as well as 1,217 posts containing no media (the text
content of which primarily related to off-site CSAM trading or grooming of minors).
From post metadata, we observed the presence of emerging content categories
including Computer-Generated CSAM (CG-CSAM) as well as Self-Generated CSAM
(SG-CSAM).

[-] Rivalarrival@lemmy.today 5 points 1 year ago

How are the authors distinguishing between posts made by actual pedophiles and posts by law enforcement agencies known to be operating honeypots?

load more comments (2 replies)

[-] godless@latte.isnot.coffee 4 points 1 year ago

Because it's talking about a report without linking to any report. That's shouting into the void at best, clickbait at worst.

[-] Sentinian@lemmy.one 4 points 1 year ago

I was able to click and access the report fine

[-] Rooki@lemmy.world 4 points 1 year ago

Because its literally nothing a normie would read through. And some thought lemmy had bad ui.

load more comments (1 replies)

[-] Rooki@lemmy.world 22 points 1 year ago* (last edited 1 year ago)

Another "SCREAM" for "BAN FEDIVERSE ITS DANGEROUS!!!!!!!". And then of course its tagged "CSAM" lol. You want a (small) website removed just accuse them of csam. And boom hoster and admins raided.

[-] Scew@lemmy.world 24 points 1 year ago

Better ban school too. I hear that's how kids find drug dealers.

[-] DmMacniel@feddit.de 6 points 1 year ago* (last edited 1 year ago)

And don't forget to ban Sugar. Because Hitler loved Sugar!

[-] keeb420@kbin.social 3 points 1 year ago

And air. I heard Stalin loved air.

[-] DmMacniel@feddit.de 2 points 1 year ago

Gods damnit. Everything needs to be banned now!

[-] FlyingSquid@lemmy.world 9 points 1 year ago

Not at all. Sensible suggestions.

load more comments (2 replies)

[-] Rottcodd@kbin.social 16 points 1 year ago

This isn't science - it's propaganda.

[-] godless@latte.isnot.coffee 9 points 1 year ago* (last edited 1 year ago)

What report?

Edit: https://purl.stanford.edu/vb515nd6874

Somehow the link in the post is broken. Trying to share it copies the link. Weird.

load more comments (7 replies)

[-] AnonTwo@kbin.social 9 points 1 year ago

Is this the same report that was brought up where it was found out Twitter has the exact same issue?

[-] Kbin_space_program@kbin.social 10 points 1 year ago

Or that Reddit has had this issue in spades.

Frankly, that the ability for the Fediverse to cut off problem servers is listed as a drawback and not an advantage is in my opinion, wrong.

[-] blazera@kbin.social 5 points 1 year ago

basically we dont know what they found, because they just looked up hashtags, and then didnt look at the results for ethics reasons. They dont even say what hashtags they looked through.

[-] Aesthesiaphilia@kbin.social 5 points 1 year ago

We do know they only found, what, 112 actual images of CP? That's a very small number. I'd say that paints us in a pretty good light, relatively.

[-] dustyData@lemmy.world 5 points 1 year ago

112 images out of 325,000 images scanned over two days, is about 0,03% So we are doing pretty well. With more moderation tools we could continue to knock out those sigmas.

[-] blazera@kbin.social 2 points 1 year ago

it says 112 instances of known CSAM. But that's based on their methodology, right, and their methodology is not actually looking at the content, it's looking at hashtags and whether google safesearch thinks it's explicit. Which Im pretty sure doesnt differentiate with what the subject of the explicitness is. It's just gonna try to detect breast or genitals I imagine.

Though they do give a few damning examples of things like actual CP trading, but also that they've been removed.

[-] Rivalarrival@lemmy.today 4 points 1 year ago

How many of those 112 instances are honeypots controlled by the FBI or another law enforcement agency?

load more comments (6 replies)

[-] moistclump@lemmy.world 3 points 1 year ago

Abstract:

The Fediverse, a decentralized social network with interconnected spaces that are each independently managed with unique rules and cultural norms, has seen a surge in popularity. Decentralization has many potential advantages for users seeking greater choice and control over their data and social preferences, but it also poses significant challenges for online trust and safety.

In this report, Stanford Internet Observatory researchers examine issues with combating child sexual exploitation on decentralized social media with new findings and recommendations to address the prevalence of child safety issues on the Fediverse.

load more comments