It's difficult right now because they are mostly manual or semi-automated.

Tribes can choose to mute the bad actors. Depending on the mute setting, the user may or may not earn any tokens. However, I heard from Marky that it isn't reliable as we are seeing some fish that got through the net.

In the screenshot above, many abusers have chosen to upvote their own comments to reduce risk of detection. It's not apparent on Steem token-based frontends due to their tiny SP, but stick out like a sore thumb on the tribal frontend.

It's kind funny really, these dickheads' behavior corresponds to the way they act before the tribes. Many were flagged by SFR or are on the Global Black List.