up vote 5 down vote favorite

We are getting a number of spam answers. Suresh is deleting them quickly but that is not a permanent solution.

I think SE system should be more intelligent and doesn't allow those spam post in the first place. The structure of these spams are very similar, so it should be easy to disallow them, but I think a more permanent solution would be better. So I suggest that:

User's with reputation 1 should not be able to post answers containing links (possibly with the exception of a white list, say links in edu domain, arxiv, citeseer, springer, doi, ...).

We can analyze the links in the posts we have got so far and check the most linked sites and then exclude them.

share
2  
You probably already know this, but just to make sure, new users (rep < 10) cannot include more than two links in a single post. – Tsuyoshi Ito Sep 30 '11 at 20:02
1  
@Tsuyoshi, I did, but thanks for reminding it. The problem is the spam answers that we get have only one link in them. – Kaveh♦ Sep 30 '11 at 21:20
1  
are these spammers human? Will a CAPTCHA do them in? – Artem Kaznatcheev Oct 3 '11 at 21:08
@Artem, I don't know, but since the format of spams are very similar I would guess so. So, yes, CAPTHCA can help. :) – Kaveh♦ Oct 4 '11 at 5:11
1  
So a solution would be an automatic CAPTCHA for any low-rep users who post answers containing links. This poses hardly any burden for legitimate users, and hopefully solves the problem. – Peter Shor Oct 13 '11 at 13:21

3 Answers

up vote 2 down vote

Why not keep answers/comments (with links) by low-rep users unpublished until a moderator/high-rep user approves? Just like it is done for edits, only with different bounds.

This would not reduce moderator's workload but would keep spam from going live while not putting off newcomers.

share
up vote 1 down vote

They have such a filter in place for comments, from what I recall. But if you prevent 1-rep users from posting links in an answer, how are they supposed to be able to post a good answer ?

share
2  
By white listing most often used sites! :) I haven't checked but I think essentially most of the links on cstheory are either to SE, MO, Wikipedia, and papers (on university websites, springer, citeseer, arxiv, eccc). We can make a short white list. Also it might be reasonable to allow links for registered users, my guess is that the spams are posted by unregistered users (you are quite fast in deleting spams so I haven't been able to verify this). – Kaveh♦ Sep 30 '11 at 21:24
@Kaveh: The list of all university web sites is definitely not short. – Jukka Suomela Oct 1 '11 at 17:34
@Jukka, pragmatically we only need to add the domain names: .edu (and possibly .ac.??). Also remember this is only for no-rep users, after one upvote a user can update the answer and add the link. I don't know if you have noticed it but recently we have started getting considerable number of spam answers. That is the problem and this is my suggestion how to deal with it, if you have a better solution suggest it. :) – Kaveh♦ Oct 1 '11 at 20:16
@Suresh, could you please check the ip address of spammers before deleting their answers. We can try asking SE to block the range if it is fixed. – Kaveh♦ Oct 1 '11 at 20:24
ah sorry. I'll do that. – Suresh Venkat♦ Oct 1 '11 at 20:56
Don’t the server admins check the IP addresses of posters of the spam posts? – Tsuyoshi Ito Oct 1 '11 at 22:48
but they don't know it's spam. at least not until I delete it – Suresh Venkat♦ Oct 2 '11 at 3:19
I do not know how moderators handle those posts. Is it different from what ordinary users do? (Ordinary users would flag them as spam.) – Tsuyoshi Ito Oct 2 '11 at 8:24
ordinary users flag the posts as SPAM and we can delete them. In these cases, the same user was posting the same spam on multiple questions, and so I deleted the user ID – Suresh Venkat♦ Oct 2 '11 at 22:34
@Suresh, if a post gets 5 spams it is automatically deleted IIRC. Moderators can also flag the posts as spam in place of deleting, but the effect is sum of flagging and deleting. (so it is better for mod to flag as spam in place of simple delete). – Kaveh♦ Oct 3 '11 at 13:56
right. but I was deleting the user, not the post. it wasn't the case of a "regular" user accidentally spamming. – Suresh Venkat♦ Oct 3 '11 at 15:23
could we put some kind of double validation, either it is on the white list, either it is not and then it should be validated by a moderator (just as editing community wiki for someone with less than 100rep). Then the white list grows with the number of validated answers? – Gopi Oct 3 '11 at 15:36
@Suresh, my guess is that SE system probably has something to deal with spammers, maybe it automatically blocks their ip, so maybe it is better to just flag them as spam and see what happens. (I am a little bit bus this week, will check it with them next week.) – Kaveh♦ Oct 4 '11 at 5:14
Just so that you are aware of what I do: If I see a spam post, then I immediately delete the user. Many of the recent posts have the same format and link to the same site, so they are probably automated. – Dave Clarke♦ Oct 4 '11 at 7:23
up vote 1 down vote

I am a little worried about false positives, but I do not have any concrete data to argue about your proposal objectively and therefore I do not object to it.

However, I would not expect too much from it. For example, it will not prevent targeted attack because there are probably some open URL redirectors on whitelisted domains which a spammer can use to post an arbitrary link.

share
yes, spammers can eventually find away to get around what we do to prevent them and then we have to find a new way to deal with that. I am not much worried about false positives, what I am suggesting is not much different from what is already there, and after one upvote the new user can add whatever links he/she wants to. I think it is unlikely that a non-spammer new-user will have trouble with this more than once. – Kaveh♦ Oct 1 '11 at 22:53
3  
@Kaveh: I might walk away from a website if I spend some time to compose an answer and it is rejected by a spam filter. Even if it is just for the first time, the first time becomes the only time. This risk already exists, but the more filters we introduce, the higher the risk becomes. As I said, I do not have any concrete data and therefore this is just a speculation, but it might be worth thinking about. – Tsuyoshi Ito Oct 1 '11 at 23:00
I understand your point. The same thing applies to not being able to comment and having more than one link in an answer. Existence of abusers makes life harder for non-abusers. Maybe SE system already has something for this situation (which doesn't kick in here because we are deleting the spam answers in place of letting users flag them as spam). I will check it with them. – Kaveh♦ Oct 1 '11 at 23:55

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged