How big a problem is chat spam these days?

[Deleted member 31926]

While hanging out on Chaturbate a while back I noticed a model's public chat was being filled with messages from "members" (probably bots) like

New cam site opening! Check my profile and get free tokens when you sign up.
​

While this was going on the model herself was, ah... she had her hands busy, couldn't easily reach the keyboard and may not even have been able to see the messages.

I've found a couple of threads on the forum talking about chat spam, but both are years old and suggest the problem was eventually solved by the Webcam hosts themselves. It seems for at least Chaturbate, this is no longer true.

Models, have you had to deal with this problem yourself recently? Is it something that annoys you? How do you respond to spammers posting messages in your room?

 

[Mila_]

It is annoying but there is always the option to mute greys. Color nicknames dont spam.

 

[FluffyGalore]

I had a guy on MFC spam loads of individual letters in my chat. It took me a while to realise he was actually trying to spell delightful not quite sure why, maybe he found me delightful lol, regardless single letters not needed and quite annoying. As for actual spam once or twice when I first started but I mute guests and basics and now it's not an issue.

 

[DahliaDee]

It has become a serious problem lately on Chaturbate. I hate muting greys, because I typically get good conversation from them, but if my mods aren't around for whatever reason it's infuriating. Once your room count passes 60 viewers or so every other message is from a spambot. It drives me bonkers, I wish Chaturbate would figure out a way to stop it

 

[Deleted member 31926]

Thanks, guys. So it sounds like

• Spam really only ever comes from basic and guest accounts (which makes sense, since having to continually fund new premium accounts would make spamming uneconomical);
• Spam is practically a non-issue on MyFreeCams as there is no disincentive for models to simply mute all non-premium accounts, cutting it off at the source; and
• Spam on Chaturbate is more of a problem as models benefit from allowing basic accounts to speak (why, I wonder; fewer users overall?) and the platform itself seems to do little to help.

I was thinking about this a few days ago and wondering whether the same technique commonly used to detect spam email would be effective in recognizing spam chat messages as well. One issue I see is the effort involved in flagging comments to train the system, which I doubt models would have the time, inclination or perhaps even opportunity to do (but then maybe their room helpers would, as @DahliaDee suggests). If this were done, though, I can imagine a tool that could step in and automatically mute a suspected spammer, giving the model a chance to review their messages and behaviour before deciding whether or not to allow the user to remain in her room. I know I would want something like that.

Anyway. Given enough time and a suitable corpus of chat messages I'd be tempted to try it and find out.

 

[DahliaDee]

The problem with this is that the spammers change their methods frequently. At first it was "Click my profile to see hot young girls!" and then models used their apps to block anyone that said "click" for example. So they changed it so the USERNAMES say "clickmyprofile1234343" and now they just use random words. No lie the other day I got a bunch of them that posted in my room saying "stocking refrigerator potato puppy". Every time measures are put in place to deter them, they come up with some new way to get around the blocks, staying one step ahead every time. I think to combat this, CB needs to make it harder to create an account - perhaps require that they use a verified email if they want to post in chats, for example.

 

[weirdbr]

• Spam is practically a non-issue on MyFreeCams as there is no disincentive for models to simply mute all non-premium accounts, cutting it off at the source; and

The problem is that spam on MFC happens differently - usually with premium accounts (possibly stolen/hijacked), which reduces its incidence, but it happens - I bet most models/members who have been around for a while remember lindasogentle - it started with a single account, then it changed accounts frequently to avoid countermeasures.. Eventually either MFC succeeded in blocking that specific spammer or they gave up.
Now, every once in a while you will see someone going from room to room spamming something - the latest two I've seen are spamming a 'video' link (it is an URL ending with '.mp4', but it actually redirects you to a site full of ads/potentially trying to steal your data - I didn't dig at the code) and the other was someone going from room to room asking for 'help' to improve their profile (which probably was a lie to get you to look at their profile, which was full of nonsense theories about models helping the CIA monitor muslims and 'identifying' such models).

I was thinking about this a few days ago and wondering whether the same technique commonly used to detect spam email would be effective in recognizing spam chat messages as well.

Baesyan filtering is nice, but it has its issues and as such is rarely used as the only signal to deal with spam - the biggest one being that users frequently pollute the training data by incorrectly flagging real content as spam (for example, user subscribes to a mailing list, then some time later decides they don't want those e-mails anymore, but instead of clicking on the unsubscribe link they click on 'flag as spam', which works for them as they never see the e-mail unless they check their spam folder). On cam sites, this could happen by models/users flagging anything from the 'mmmmm bb' folks (who aren't necessarily spammers, but could learn how to use words) to people they consider to be too verbose/annoying on top of the real spam, at which point real users would get muted/banned.

 

[justjoinedtopost]

the other was someone going from room to room asking for 'help' to improve their profile (which probably was a lie to get you to look at their profile, which was full of nonsense theories about models helping the CIA monitor muslims and 'identifying' such models).

Saw it. Nutty shit.

 

[Deleted member 31926]

Thanks, @DahliaDee and @weirdbr. That gives me a broader perspective on the problem. I can see now Bayesian filtering definitely wouldn't be enough on its own, especially if spammers are using random words in their messages each time. And while I can think of other signals (like frequency of posts) that might help give a spammer away, it's clear the kind of information you'd need means this could only really be solved effectively at the platform level anyway. Pity Chaturbate isn't trying a bit harder!

 

[DahliaDee]

Thanks, @DahliaDee and @weirdbr. That gives me a broader perspective on the problem. I can see now Bayesian filtering definitely wouldn't be enough on its own, especially if spammers are using random words in their messages each time. And while I can think of other signals (like frequency of posts) that might help give a spammer away, it's clear the kind of information you'd need means this could only really be solved effectively at the platform level anyway. Pity Chaturbate isn't trying a bit harder!

It really is a pity. It's a huge problem on there and it's really degrading the quality of the site.

 

[Marissa Frost]

I miss LindaSoGentle