Page MenuHomePhabricator

AbuseFilter integration for JADE
Closed, ResolvedPublic


Judgments should be sent through AbuseFilter, for which we need to implement at least the AbuseFilter-contentToString hook to flatten JSON, maybe reducing to just the freeform text values, or including quantitative judgment values in the hope of catching "marblecake"-style attacks.

Event Timeline

Do you mean that the judgements should be subject to abuse filters? (E.g. if they contain profanity, it should be possible to block saving that judgement using an abuse filter that matches the profanity and prevents the action?)

@Huji Yes, exactly. We're hoping to create a new namespace "Judgment" and want to take every possible precaution against creating additional patrolling workload. It probably can't be helped in the end, but AbuseFilter and ORES flagging will help with reduce work by catching the low-hanging fruit.

Harej triaged this task as High priority.Aug 8 2018, 4:48 PM
Harej moved this task from Inbox to Feature Requests on the Jade board.

Change 464716 had a related patch set uploaded (by Awight; owner: Awight):
[mediawiki/extensions/JADE@master] AbuseFilter integration test

awight moved this task from Backlog to Review on the Machine-Learning-Team (Active Tasks) board.

Looks like text filtering will work out of the box.

Change 464716 merged by jenkins-bot:
[mediawiki/extensions/JADE@master] AbuseFilter integration test