loading...
Cover image for Dev.to: How to stop the spammers?

Dev.to: How to stop the spammers?

jorgecc profile image Jorge Castro ・1 min read

It is just a case study.

Let's say we run the administration of dev.to, and we want to stop those spammers?.

What is the best method to do that?.

  • Blacklist words. It works but limitedly. It also could ban legit users.
  • A twitter or a GitHub account. It is not working.
  • Technically, it could use some AI deep-scanning or whatever it is called. However, it is an expensive operation, and it is also ineffective (spammers learn how to bypass it in a snap).
  • Manual control. It is a hell of an expensive.

What we know about it.

Those spammers are tricky, and they know how to bypass automatic restrictions. For example, they upvote their own posts. They also fill with lots of text, apparently to bypass some restrictions. And some of them also filled with a c&p test (from somewhere else), but finally, they spam with links.

Also, it is funny, but they use stock photos.

However, I don't understand them. Their methods are unpractical. They spend so much time to spam it.

So, what is the best alternative to stop spammers without breaking the piggy bank?

Posted on by:

jorgecc profile

Jorge Castro

@jorgecc

You are free to believe in whatever you want to, me too. So, stop preaching your religion, politics, or belief. Do you have facts? Then I will listen. Do you have a personal belief? Sorry but no.

Discussion

markdown guide
 

Im part of the moderation team here, and I agree it's a real issue.

The current process for volunteer mods is to flag posts to be evaluated by a staff member when they are next available. This does no remove posts, the most we can do is a slight nudge on how likely they are to end up in your feed.

A pattern I notice, is that most spammers tend to have newly created Twitter or Github accounts with very little or no content. We could potentially use their API's to get the account age/activity and create filtering on that, however this does run the risk of excluding genuine users.

Personally I would like to see a way of having users flag potential spam content/accounts for moderators and staff to see as a priority in moderation feeds. This mixed with automatic flagging with a system similar to detailed above, would lead to much less spam imo.

This is a topic we need to take time to investigate and find the right solution for DEV. The light touch to moderation is one of this platforms main reasons for growing and retaining viewers and contributors.

If you have any potential solutions, I would encourage you to raise an issue or create a branch and PR it over on DEV's GitHub page

I don't know if any of the staff would be interested in a public discussion about this?

ben image

jess image

molly_struve image

thepracticaldev image
 

Also part of the mod team! I was most surprised to see how similar spammers tactics are. Between the title, tags, and account name, I can often tell pretty quickly if its a spam account.

I believe anyone can use the "Report Abuse" button link at the bottom of a post - there is an option for spam or copyright. Additionally, the "..." menu on a user profile also allows you to Report Abuse, same option for spam. Sure, its not super automated, but with enough data these accounts and posts are similar enough we can build something ML based in the future.

 

Wait, what spam are we talking about? I've seen a lot of posts I personally find uninteresting but not spam... might just have missed it.

 

Mostly comments that looks like "wow nice article I really enjoyed it and I'm interested in %topic name%. %random link% wow thanks for the article!". I've seen a bunch of these comments but I always report them.

 

I've flagged a few.

Often the ones I see are weirdly long content, with poor formatting, that read like they've been generated by a Lorem Ipsum type generator. Also, with a lot of links. Often but not always about topics that aren't related to typical DEV content.

The posts may have 'hearts' and comments, which give it some credibility but these are probably from other spam accounts.