Concept:
Twitter spam mitigation via SURBL applied to young Follow ratio inequity accounts
Synopsis:
Twitter should use SURBL checking against all published URL pushed by early life accounts for 90 days where a Follower to Following ratio is greater than ~100:1 as part of building a minor reputation score.
Details:
Based on what I have noticed since being on the service, there are a lot of spammer and developer possibilities right now. As Twitter operations and devel hooks have cranked down on certain areas there has been gravitation to other exploits or testing of the system.
A typical spammer account has the following characteristics:
1) < 10 followers with auto follow NOT selected
2) following +1000
3) +1 post with a obvious SURBL hit i.e. a spamvertized URL
4) Less than 90 days old
Considering how Blogger has dealt with these issues and other form based services, Twitter should use SURBL filters.
Proliferation:
Of course, obfuscation via shorter URL services might crop up overnight but these are also easily tracked in the XMPP aggregate to a manageable level. Consider that 100 new accounts of the form {girlname}{bigint} are likely not real people, are bots, and bots bent on spamvertising a single URL.
Secondarily, as with any emerging reputation system there will be shill creation and shill for hire if a follower reputation is created based on auto-follow classification.
Third, it is likely that any internal algorithm will eventually be gamed as the corpus of user creation maps to larger and more automated applications. 90 days is a sliding scale that will require tuning.
Afterword and Suggestions
While I appreciate the effect of growing net new accounts for Twitter via bot techniques of spammers -- it starts to feel like Blogger. I'd like to suggest that Twitter consider
a) SURBL implementation guided by community inputs and tempered by anti-spam leaders
b) shortly thereafter ratify or endorse a SURBL compliant (stated) URL shortening service
Conclusion
Enforcing reputation and checks against SURBL removes a key vector being used by these namespace and follow approaches by the current run of spammers on Twitter.
Reply to this idea