Abuse of multiple accounts

5 views
Skip to first unread message

Nick Arnett

unread,
May 8, 2009, 11:00:37 AM5/8/09
to twitter-deve...@googlegroups.com
I knew this would happen... one person with a bunch of accounts has managed to spam my social network analysis:


In this case, it is very obviously the same person, since she is using the same picture for every account and only slight variations of her real name.

I can detect some of this by seeing real names that correlate to multiple identical tweets... Curious if anybody else has thoughts on ways to identify this sort of abuse.  Perhaps if the API told us what percentage of people block each user?

Just noticed that most of her profiles have the same home page URL, so that's a strong clue... and most of her tweets contain the same URL.

I'm sure that Twitter's fraud group uses some sort of scoring system... any chance that any of that data could be shared in the API to help automated systems avoid retweeting spam?

Nick

Matt Sanford

unread,
May 8, 2009, 11:32:41 AM5/8/09
to twitter-deve...@googlegroups.com
Hi there,

    We do have a slew of reports and tools for our abuse team looking at blocking, duplicates and some "secret sauce" to find bad accounts. I'll pass this on and see if it wasn't caught for some reason or is in the process of being handled. As far as sharing our data it via the API we have no plans to do that. The issue isn't showing the data to friends, it's showing it to enemies. I think the development community could probably come up with some cool analysis on this, but so could the spammers. If you show your opponent all of your cards they will raise the stakes.

Thanks;
 – Matt Sanford / @mzsanford
     Twitter Dev

Nick Arnett

unread,
May 8, 2009, 11:44:30 AM5/8/09
to twitter-deve...@googlegroups.com
On Fri, May 8, 2009 at 8:32 AM, Matt Sanford <ma...@twitter.com> wrote:
Hi there,

    We do have a slew of reports and tools for our abuse team looking at blocking, duplicates and some "secret sauce" to find bad accounts. I'll pass this on and see if it wasn't caught for some reason or is in the process of being handled. As far as sharing our data it via the API we have no plans to do that. The issue isn't showing the data to friends, it's showing it to enemies. I think the development community could probably come up with some cool analysis on this, but so could the spammers. If you show your opponent all of your cards they will raise the stakes.

I certainly understand that, but I was thinking more of a score, rather than any information about what's behind the score, to use as evidential logic. I can see why it is safer and easier to just keep it all behind the scenes until and unless the account is shut down.

Any chance of sharing the percentage of people who have blocked each user?  That's feedback from your users, after all, and thus somewhat "belongs" to the community.  (There's probably a huge hole in that argument somewhere, but I'm not going to think about it).

Nick

Patrick Burrows

unread,
May 8, 2009, 1:27:47 PM5/8/09
to twitter-deve...@googlegroups.com

Interesting. Most of her tweets seem to be pretty random and meaningless as well. (though, I suppose the same could be said for many legitimate people.)

 

I don’t imagine this is the sort of account Twitter would pick up on and ban, either (that was my first thought – just wait for Twitter to ban it.)

 

--

Patrick Burrows

http://Categorical.ly

@Categorically

Doug Williams

unread,
May 8, 2009, 1:49:13 PM5/8/09
to twitter-deve...@googlegroups.com
Actually this set of accounts are prime targets to eventually get swept up by one of our automated spam algorithms. This data (a spam score) isn't made public in large part because the code that performs the science is separate from the main twitter.com codebase. Additionally, we don't want to reveal any secrets on how to circumvent our analysis.

If you feel that someone is a spammer, please dm or @reply @spam (e.g. @spam @WealthWizz) to help in The Fight Against Crime (tm).

Thanks,
Doug
--

Doug Williams
Twitter Platform Support
http://twitter.com/dougw

Nick Arnett

unread,
May 8, 2009, 2:30:57 PM5/8/09
to twitter-deve...@googlegroups.com
On Fri, May 8, 2009 at 10:49 AM, Doug Williams <do...@twitter.com> wrote:
Actually this set of accounts are prime targets to eventually get swept up by one of our automated spam algorithms. 

That's good to hear.  I'm going to wait and see how often this happens before I start working on new code to detect it myself.

Nick 

Doug Williams

unread,
May 8, 2009, 2:56:58 PM5/8/09
to twitter-deve...@googlegroups.com
Nick,
We have a "chief scientist" in house who actually manages all of these algorithms. It is his job to determine how to spot spam and sketchy users (his words) through the data. I'm sure you can understand why we cannot share this part of our secret sauce openly.

Also, if there are isolated incidents of spam or abuse that you want to report, you can always send an @reply to @dougw and I can take care of them on my own.

Cheers,

Doug
--

Doug Williams
Twitter Platform Support
http://twitter.com/dougw



Nick Arnett

unread,
May 8, 2009, 5:05:34 PM5/8/09
to twitter-deve...@googlegroups.com
On Fri, May 8, 2009 at 11:56 AM, Doug Williams <do...@twitter.com> wrote:
Nick,
We have a "chief scientist" in house who actually manages all of these algorithms. It is his job to determine how to spot spam and sketchy users (his words) through the data. I'm sure you can understand why we cannot share this part of our secret sauce openly.

Really, I wasn't asking for algorithms... I was hoping for scores, but that's okay.
 

Also, if there are isolated incidents of spam or abuse that you want to report, you can always send an @reply to @dougw and I can take care of them on my own.

As long as they are isolated, I'm not going to worry much about 'em.  ;-)

Given that this was the first one I've noticed in months, it seems that you guys are doing a good job.

Nick 
Reply all
Reply to author
Forward
0 new messages