On Jan 8, 2008, at 10:43 PM, Steve Bennett wrote:
IMHO we should be putting more effort into brilliant vandalism detection and rollback
This area is ripe for exploration. Has anyone looked into "Summer of Code" type projects for this sort of thing? The signatures for the great majority of vandalism are not difficult to understand.
I would guess that over 50% of vandalism that I revert would be matched by these rules:
* Two to three IP edits in short succession * High likelihood that the IP has never edited before * The last edit of the series is not vandalism * The "last - 1" edit has a Lupin regex word * The "last - 1" edit is in a different section so as to not appear in the diff
I don't believe ClueBot handles this case and my Perl is teh suk... any volunteers?
--Noah--