Sysadmin by day, developer by night

Simple solution for you Twitter, heck, give me access to the resources where I can tap into the stream and I’ll build it for you.

- Scan tweets as they come in for url.
- If the tweet has a url in it, dump into a db
- Periodically (every 60 minutes) run a report on the db

Report would:

Pull each url from a tweet
Check against cache
Scan it against something like the Google Safe Browsing API
Cache the url and result for future lookups

If user sends X amount of non-safe links in X amount of time, flag them for review.

Flagged accounts are then reviewed by real people. The initial few days/weeks might take a lot of time, hire some temps. When you’re past the initial influx, it should settle down to something an employee or 2 can keep up with while doing other functions.


Edit:

OK, looks like that idea has some flaws, and would be more resource intensive than necessary. I tested some urls from a current spamming account that recently spammed me. None of them are showing up in Google Safe Browsing list.

So, plan B.

Well, statuses/links exists as a stream point, not generally available. But that would cut down on filtering whether a tweet includes a link or not.

So for each account that posts a link, perform some basic tweet history analysis

More than X tweets with links per day, drop them in the report bucket
Fire up the reporting engine periodically to check all their tweets againsts a spam analysis.

Criteria could be something like

x points for posting the same link via multiple tweets all with replies

Eventually you start to cache known bad links so you start to identify trouble quicker. Of course spam links will change fairly quickly. Pretty complex analysis engine for sure.


OR…


If you have less than X followers and your account is less than X seconds old you can only make X amount of replies per day.

“We’re sorry, your account is fresh and in an effort to better serve our existing customers we encourage you to become more engaged with your Twitter activities before sending more messages directly to individuals. Thanks”

Make it smart, give points back for replies back from members who meet a criteria of being considered members in good standing. Make sure the activity is monitored so that you can’t create an account, wait a week, then start spamming.

Amount of followers might be a good metric, I’ve noticed the last few accounts that spammed me had very few, if any, followers.

  1. sp0rus reblogged this from joerussbowman
  2. joerussbowman posted this
blog comments powered by Disqus
Technorati Profile