|
I've read this post too late, just edited some of them. But if you think we should not, I'll stop it. It was the idea of one other member, I though it was not a bad one.
What do you think about the following problem:
When and if we remove all posts of all offenders, there is no evidence of spam left. A lot of offender's accounts was removed due to member's abuse reports, but I cannot be sure that all members will trust the report if there is no evidence of spamming. Unfortunately, sometimes we can see such thing as "hatred voting". What if somebody reports a good member as one of spammers? Even if you have some spamming evidence in the database (I don't know), the members don't see it… What do you think?
And please see my next post, at the moment of writing, on top of list. I suggest you start to ban some user names permanently, at least some of them.
Thank you. Happy New Year!
SA—Sergey A Kryukov
|
|
|
|
|
I understand your concern regarding members being nuked by hatred voting, but if messages are left in a viewable state then the spammers get what they are after: listings on Google and Bing.
The guys who are jumping on these spammers all seem a very sensible lot. Let's see if there's a problem with hatred voting to close accounts before we modify the approach to fix a problem that may not be there.
I've also incapacitated the Sema account.
cheers,
Chris Maunder
The Code Project | Co-founder
Microsoft C++ MVP
|
|
|
|
|
Hi Chris.
I understand that. That explains why immediate removal of spam posts is a priority. It's convincing. My concern was not hatred voting itself. My concern is possible false negative/positive in removing spammers' accounts. This whole spamming problems needs thinking.
Thank you,
—SASergey A Kryukov
|
|
|
|
|
Sergey Alexandrovich Kryukov wrote: This whole spamming problems needs thinking
It does. Every time I try and think of something automated I always seem to fall back to a point where a human check is needed so we don't nuke accounts. As soon as human intervention is needed the whole thing seems to collapse down to simply having keen-eyed members nuke the messages.
Initially I had it so that accounts were automatically nuked after a certain number of spam messages were found, but this definitely caused false-positives, so I reverted back to separating the flagging of the message and the flagging of the accont. I do have some automatic checks on new posts, with some basic checks to ensure the posts of regular members aren't checked, but even then we've had cases where accounts are compromised and used for spamming purposes.
A foolproof algorithm, or even workflow, would be nice!
cheers,
Chris Maunder
The Code Project | Co-founder
Microsoft C++ MVP
|
|
|
|
|
I agree. My thinking was: human check up is always unavoidable, so the solution lies in automation of the analysis/detection and easy and fast application of appropriate measures.
By a simple example: right now, the distance between seeing the first spam artifact and just reporting is too long: one needs to chain all posts related to the spotted post: see the accounts, all posts from the owner of the accounts, all "answers", accounts of the authors of the "answers", and so on, until one collects the complete closure of the initial set of artifacts. This is what I really do, sometimes.
This closure could have only the initially spotted post and offender's account, but there are cases when it is much bigger than that. If one deletes the posts prematurely, the problem of collecting the full set will not be solved, because some of the URLs are lost. At the same time, solving just this part automatically is simple enough. We can have the solution where only the initial spotting of the offending post is required, and all the related information is collected automatically and presented in one place without duplicates. It may create false positive (an accounts of naive members responding to spammer's post, for example  ), which can be excluded by a human on the last step when a final decision to remove posts and account is made.
I did not even discuss the possibility of Bayesian filters, which population is be human-driven. The filtering itself gives you false positives and false negatives. False negatives (non-detected spam) could be compensated by members; false positives — by human intervention on the last stage of the final decision.
This is what I think. If can be refined and further detailed and will need especially thorough planning, but here is the thing: I'm afraid to say that not doing this development may grow more expensive than doing. See what's going on?
Thank you,
—SASergey A Kryukov
|
|
|
|
|
|
Chris,
A new idea recently came to my mind. I quickly composed a short Tips/Tricks article with some initial proposal. Will you please review it and tell me what do you think. Certainly, this is some job, but I seriously think that it may turn more affordable then our current suffering from spam and disproportional effort, if you consider total losses in time and nerve.
Besides, the implementation may set new model of spam protection for a site.
We can consider further detail and possible variants of the idea.
Please see:
http://www.codeproject.com/Tips/519762/A-Plan-for-Spam[^].
Thank you for your attention for this matter,
—SASergey A Kryukov
|
|
|
|
|
I assume you are asking us not to edit the posts so you can tune your anti-spam automation. In any case I am fine with that as I did not care for my efforts being as visible in QA as they were the other day (you never know if they come up with a way to retaliate).
I supported the idea of trashing the links in the posts because it seemed to keep the view count down compared to questions where they were left untouched. You mentioned in a previous post to SAK, that the spammers get their way when we leave the posts in a viewable state - they show up in Google and Bing searches.
Unfortunately, deleted posts also show up and for some of them you can bring up a cached view (try entering this in a Google search: ufc codeproject). I am not an expert on how long it takes from a question is posted until the page gets cached or if the spammers even get any business from cached, deleted posts so I would appreciate any insight on that if anybody knows.
BTW, I went as far as editing some of the comments with spam links before I deleted the comments. I did that because I have noticed that even after the comment is deleted and the associated question or answer is deleted, the comment is still visible through the member's profile page and I did not know if that shows up in a Google search.
Soren Madsen
"When you don't know what you're doing it's best to do it quickly" - Jase #DuckDynasty
|
|
|
|
|
Chris,
Please, what's going on? It looks like you personally blocked Sema Khan name from using it as a member name and/or friendly URL name, but he is re-appearing again and again:
http://www.codeproject.com/script/Membership/View.aspx?mid=9743696[^].
I already sent you a note on that. Did you miss something?
Thank you,
—SASergey A Kryukov
|
|
|
|
|
What I did was reactivate an account of his, change the password so he couldn't use, and let it sit. That way he couldn't reuse it since the name was already in use.
It was, however, quickly reported out of existence.
However, it's a moot point: blocking a name will not stop these guys for a second.
cheers,
Chris Maunder
The Code Project | Co-founder
Microsoft C++ MVP
|
|
|
|
|
How come it won't stop them? It won't stop from creating an account under a different name, but under exact same name... so many times... Apparently, it's possible to block forever...
—SASergey A Kryukov
|
|
|
|
|
Message[^]
User[^]
Bob Dole The internet is a great way to get on the net.
 2.0.82.7292 SP6a
|
|
|
|
|
The message is removed. What wrong was with it?
—SASergey A Kryukov
|
|
|
|
|
It was completely incoherent and looked somewhat like what CSS used to post in the back room.
Bob Dole The internet is a great way to get on the net.
2.0.82.7292 SP6a
|
|
|
|
|
|
|
|
|
http://www.codeproject.com/Members/tvfanatics[^],
http://www.codeproject.com/Members/Taylor-Goslow[^],
http://www.codeproject.com/Members/Tanalr[^],
http://www.codeproject.com/Members/ufc366[^],
http://www.codeproject.com/Members/ufc366[^],
http://www.codeproject.com/Members/fkasodsad[^],
http://www.codeproject.com/Members/ryuuseixandee[^],
http://www.codeproject.com/Members/sajekale[^],
http://www.codeproject.com/Members/Lili-Gomes[^],
http://www.codeproject.com/Members/bluerecoder[^],
http://www.codeproject.com/Members/mnrsk[^],
http://www.codeproject.com/script/Membership/View.aspx?mid=9721078[^],
… … … … to be continued …
Survived after yesterday's attack (or re-created accounts):
http://www.codeproject.com/Members/Sema-Khan[^],
http://www.codeproject.com/Members/balotpinoy[^].
—SASergey A Kryukov
modified 31-Dec-12 20:56pm.
|
|
|
|
|
|
Agree; now, can you come up with the design of such Wunderwaffe? 
—SASergey A Kryukov
|
|
|
|
|
|
The main problem is the location of the target. How about that. Civilians should not be hurt.
—SASergey A Kryukov
|
|
|
|
|
|
That's a better idea. You are moving in right direction...
—SASergey A Kryukov
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
