News for hackers. Stuff that matters.
Re:Possible routes for distributed anti-abuse systems by yummyfajitas on Wednesday April 05, 2017 @05:20AM (#14040805)
There is no abuse problem on the platform, or rather any abuse problem on the platform is already a solved problem via existing techniques (e.g. standard ML classifiers).

Consider SMTP, which is a true peer-to-peer system. In SMTP, abuse is a solved problem - spam filters work great, and 33mail/mailcatch (disposable email) improves things tremendously. At most 1-2 recruiters/SAAS offering marketing automation/etc slip through my filters each week, forcing me to expend about 3-5 seconds/week clicking "report spam".

The "abuse" problem on twitter is similarly not really a filtering issue; if it were, then it could be easily solved by a moderately better twitter client or switching off notifications of @mentions.

No, the actual problem people face is that they want to stop others from speaking, and label the speech they wish to stop "abuse". What's scary to most (including me) about Twitter is that an angry mob may form and decide they hate me, and then use Twitter to coordinate real life actions against me.

That's not a problem of abuse on the platform; that's a problem that the platform enables people to speak among themselves in ways I dislike. Lets be clear that this is the problem some folks wish to solve. It's only by identifying the problem that we can properly solve it, or decide if it's a problem that should be solved.

Re:Possible routes for distributed anti-abuse systems by pjc50 on Wednesday April 05, 2017 @06:17AM (#14040999)
> What's scary to most (including me) about Twitter is that an angry mob may form and decide they hate me, and then use Twitter to coordinate real life actions against me.

There's multiple things going on here. We've seen use of communications channel A to organise harrasment on channel B - which is hard for B to stop ("brigading", gamergate). Gamergate escalated into real-life death threats against its targets. We've seen on-platform organisation of harrasment of people on and off the platform (this is ultimately what got Milo banned from twitter). We've also seen "news" organisations get into finger-pointing and condemnation of random citizens (e.g. Daily Mail outing a teacher as trans, resulting in her suicide).

We've seen the legal system get involved clumsily. There's seemingly no middle ground between "joke in poor taste" and "arrested for bomb threats" (#twitterjoketrial). Occasionally people resort to libel law, and even more occasionally the less powerful, famous and obnoxious person wins (Katie Hopkins vs. Jack Monroe).

People are I hope aware of the ability of abuse reporting systems to themselves become channels of abuse. Facebook's real names policy is exploited by anti-trans campaigners to force people off the platform, for example. Anyone should be aware of the risks of automated threshold systems being abused: if all you have to do is get 100 accounts to press "report abuse", that will be abused very quickly.

Local standards also present problems. Do we really want to go along with e.g. Pakistan arresting people for blasphemy?

Re:Possible routes for distributed anti-abuse systems by paroneayea on Wednesday April 05, 2017 @09:27AM (#14041902)
People are I hope aware of the ability of abuse reporting systems to themselves become channels of abuse. Facebook's real names policy is exploited by anti-trans campaigners to force people off the platform, for example. Anyone should be aware of the risks of automated threshold systems being abused: if all you have to do is get 100 accounts to press "report abuse", that will be abused very quickly.

Local standards also present problems. Do we really want to go along with e.g. Pakistan arresting people for blasphemy?

It's definitely true that anti-abuse systems can be themselves abused, though most of the systems that you're talking about are partly due to the anti-abuse systems being centralized, right? I also see a lot of comments here along the lines of "but that's censorship!" But the article is discussing decentralized anti-abuse systems which allow individuals to set up their own opt-in filters which apply to themselves and their communities (which means different people might have different filters). Do you think that's different?

Re:Possible routes for distributed anti-abuse systems by pjc50 on Wednesday April 05, 2017 @10:27AM (#14042286)
Filters deal with the situation where A is sending to B something that B doesn't want to recieve.

The situation where A is sending to B something that's harmful to C cannot be dealt with by C's filtering and can only be addressed at a higher level in the system.

Those are the technical distinctions, but there's a lot of possible things covered by the second case: leaked nudes, lynch mob organisation, slander, leaked intelligence, compromised party documents, names of human rights activists being leaked to secret police, copyright infringement, child porn, fake news, real news in fake states, allegations that invitations to pizza are evidence of child porn, and so on.

Re:Possible routes for distributed anti-abuse systems by yummyfajitas on Wednesday April 05, 2017 @10:34AM (#14042342)
Other things covered by this case include leaks identifying corrupt behaviors, allegations of sexual harassment such as Susan fowlers, and evidence of human rights violations that the authorities wish suppressed.

Or various right wing ideas now softly suppressed on Twitter/Reddit.

Re:Possible routes for distributed anti-abuse systems by someguydave on Wednesday April 05, 2017 @11:23PM (#14047686)
>The situation where A is sending to B something that's harmful to C cannot be dealt with by C's filtering and can only be addressed at a higher level in the system.

Huh? If C knows A's public key and the content is signed, why can't C filter A's content?

Re:Possible routes for distributed anti-abuse systems by pjc50 on Thursday April 06, 2017 @04:28AM (#14048868)
Content is basically never signed, and I'm talking about situations where the content isn't intended for or sent to C.
Re:Possible routes for distributed anti-abuse systems by tnone on Wednesday April 05, 2017 @08:09AM (#14041452)
Gamergate, never allowed to speak for itself, never defined, never counted, but somehow, the people who invoke it as a boogieman know exactly who did what and why.

That should tell you what was going on, especially when what set it off was an abuse victim warning that his abuser was capable of anything for self aggrandizement.

Candace Owens / Social Autopsy found that out when she started getting threatening and hateful emails from "gamers" on an address she'd only given out to progressive "anti abuse" activists.

Re:Possible routes for distributed anti-abuse systems by milesrout on Wednesday April 05, 2017 @08:01AM (#14041412)
>"brigading", gamergate). Gamergate escalated into real-life death threats against its targets.

No it didn't. It really didn't. No more than Twitter normally is. 'I hope you die' and 'I hope you get shot' aren't death threats, they're horrible things to say. There's a difference.

Re:Possible routes for distributed anti-abuse systems by pjc50 on Wednesday April 05, 2017 @08:59AM (#14041741)
http://www.businessinsider.com/brianna-wu-harassed-twitter-2... (Trigger warning: contains mildly redacted actual death threats)

(mind you, the whole question of #twitterjoketrial was whether "Crap! Robin Hood airport is closed. You've got a week and a bit to get your shit together otherwise I'm blowing the airport sky high!!" constituted an offence against the Communications Act 2003 by being of a "menacing" character.)

Re:Possible routes for distributed anti-abuse systems by mafribe on Wednesday April 05, 2017 @01:05PM (#14043622)
What reason is there to believe that Wu (aka John Walker Flynt) and the businessinsider report are credible sources? There are financial interest at play here that I'm not being given sufficiently strong reasons to ignore.

https://en.wikipedia.org/wiki/Confirmation_bias

Re:Possible routes for distributed anti-abuse systems by milesrout on Wednesday April 05, 2017 @05:55PM (#14046003)
But that's an isolated incident. Those aren't common, and even if they were common, Twitter is great about reporting those people to the police.
Re:Possible routes for distributed anti-abuse systems by M_Grey on Wednesday April 05, 2017 @09:09AM (#14041798)
You make it sound like speech hasn't always been a contest; nothing about the freedom from government censorship implies anything else either. All you get with "free speech", is the notion that the government can't thumb the scales with their immense power and influence (having the ability to create and enforce laws).

The problem does have a way to solve itself, but that probably involves the model we see here on HN, which admittedly is preferable to the model of a locale such as Reddit. That said, each platform is going to want to address the problem on a faster timeline, because the problem directly hits their bottom line.

Re:Possible routes for distributed anti-abuse systems by yummyfajitas on Wednesday April 05, 2017 @10:28AM (#14042298)
I have no idea why you bring up the government - no one besides you is discussing government censorship. Are you attempting to steer the conversation towards a discussion of your libertarian principles?
Re:Possible routes for distributed anti-abuse systems by nateberkopec on Wednesday April 05, 2017 @09:31AM (#14041926)
Twitter took years to deploy notification syncing across clients (if I read a message on mobile, it gets marked as read on desktop). Is it really so hard to believe that deploying a "spam filter" for hate speech is difficult for them?

I agree that the technology exists (and plenty of other solutions mentioned in the article). But saying that because the technology exists and hasn't been deployed therefore there is no abuse problem is a hilariously bad non sequitir.

Re:Possible routes for distributed anti-abuse systems by yummyfajitas on Wednesday April 05, 2017 @10:29AM (#14042303)
It may be a problem on Twitter, but there is no reason to believe it's a problem on p2p social networks.