What can we do to prevent posting of spam links? Have a reputation limit, minimum number of posts? Anything else?
Links show in posts, user profile "about" and the "website" sections.
Is the problem spambots or human spammers? Doesn't Askbot have a Captcha for certain reputation levels?
Also, perhaps you could disable url conversion for a certain reputation level.
This could be applied to both posts and profiles.
A more complex solution might be to have urls hidden in low rep posts (and profiles) pending approval from a moderator.
It sounds right to prevent URL conversion to links depending on the reputation level.
I like the idea of preventing URL conversion, at least until moderator review. But I suspect spammers would still post plain text links judging from my wiki and Trac systems. My "more complex" solution would be a category of post/comment that was only visible to the poster and moderators (or members with high karma) if it matched a regex blacklist, until explicitly made globally visible.
Disabling links for low rep users, esp in their profiles. In particular, rewrite the URL with spaces so it is human not computer readable. Also, nofollow links for medium rep users. This would be great!
I can tell you for a fact that spammers are googling for OSQA and askbot footer text, so they can drop links. Astrong antispam measure like I propose would be great!
A compilation of ideas:
its perfectly fine, not a problem at all. :)
How can I force low reputation users to solve captcha when making posts? Is this baked into a new version of askbot?
I just went through an exercise of removing the spam from the profile sections. Sneaky spammers putting links to really unrelated websites. Here is the query that i used
update auth_user set about = '' where reputation = 1 and about != '';
idea What if we disallow users to edit the "about" and "website" sections of their profile unless they have greater than 1 (or configurable) reputation.
Some alternative/supplementary ideas:
Things that can have links like the personal website or profile description when edited could come to the attention of the admin.
At the moment, I do a regular manual check myself of the profile descriptions to see and ban people posting spam in there.
Banning a user could also delete all of their posts. But that is less of an issue right now.
that's a one spam-busting query!:)
Please also log the IP address used to create the account so that an admin can review for clues to see if the user is posting where they say they're posting from.
For example, I got a one-line question on my hosted site asking about health insurance in NJ. This is borderline but probably off-topic for an expat forum. I posted a comment asking him to clarify, but I never got a response, so I closed and eventually deleted the question.
His account fits one pattern that makes me suspect it was a spam bot:
However, I can't be sure.
If I knew the IP address that created the account, I could be more confident in deciding whether this is a legit account.
This would also be useful for detecting sockpuppet behaviors.
Great suggestion! I think to have the IP of the user is a must have.
One suggestion would be to allow only openID services for login and not local ones, as discussed here. This way you can eliminate spammers, because users will have to give a valid account.
Now, for those who do give valid accounts but are still spammers and post questions with offensive content, there is always the community to flag and vote them down (I think afrer few flags, post is automatically removed from homepage, right?).
Last, for those who add a spam URL in their profile, I think you should just require some 50 points of karma to allow adding personal website or profile description.
We could hide the url by showing link text instead of the actual link and tell the user that it is hidden until their reputation reaches some threshold.
Unfortunately spammers register "throw away" OpenID accounts.
I'd like that every link has changed in a default text ( for example LINK SUGGESTED ) with mouseover event that show the link inside.
In this way i prevent the reading of classic spam url and the automatic click that some users do without thinking ( especially to shorted url ).
The best solution would be that this link excpet to be changed in the way that i suggest is that an internal antispam system and analyzer of dangerous url identify the url suggested showing it in :
red color if dangeorous
yellow if not sure but ..
green if ok
I know that is a big work but when we suggest a solution why not to suggest the best way?
Obviously if in the comment or answer or question will be also other kind of flag and not only OFFENSIVE but also SPAM could be useful to categorize it with other type of informations.
+1 interesting idea. Would be a major plus if there was already a content evaluation service with an API we could hook into to make the red/yellow/green determination (e.g., Net Nanny or OpenDNS content filters).
Some antispam measures taken for Ask-Sage (an askbot-powered Q&A site for the Sage Mathematical Software System) recently (2016-01) were discussed in this sage-askbot-devel thread. Hopefully they can be useful more widely.
Comments