I run @sadkitten so you don't have to - dealing with massive amounts of spam

Spam makes @sadkitten sad ... 😿

Recently I noticed the new @spaminator account publishing posts claiming that thousands of accounts are engaging in bot net spam scams. Not only claiming it, but they also had the usernames of all these accounts! It seemed pretty crazy there could be this many.

I thought it might be cool to train an AI to discover these accounts. Before I could get into that @transisto got me in touch with @patrice, the person who runs @spaminator, and after a conversation also including @pfunk I decided I would create a bot to counter the rewards on these comment farming schemes for the accounts already identified.

So that's what we put @sadkitten to work doing.

What is the scheme?

Some of the accounts make comments on their own posts or other random posts with a predefined dictionary of comments. One class of these accounts uses generic "nice post" type comments (example) and some others use random nouns (example). These are up voted by a large number of the other accounts (example) which invariably have a very small amount of SP. I'm not sure if this is just the amount delegated by Steemit on account creation or if they also include further delegation.

So this is flagrant spam with the purpose of extracting rewards from the reward pool. Per comment the amount is small but by doing it a lot it really adds up.

We know of around 5,500 accounts doing this at the moment.

What to we do to counter it?

Firstly I looked through a lot of the accounts to double check. There are a lot of them so I didn't check every single one, but I did look at a random sampling of several hundreds. @patrice had checked them first so this was just really to convince myself. I was convinced.

I created a bot which tries to counter the rewards on these posts, most of which are comments. @sadkitten has almost 150,000 SP delegated from @transisto so this was quite a lot to counter these small rewards. You can see for yourself on steemd for steemdb, but the flags were almost always less than 1%.

I'm proud to say that so far the bot has scanned through many days of these spam posts and used up from 100% down to 28.52% voting power! That's a huge amount and gives you an indication of how numerous the posts are. Over 7,500 flags were made in the last 24 hours.

What now?

We also added some accounts from @sherlockholmes, which both myself and @patrice checked and their list was legit. So we will continue to add spam accounts and run the bot to counter these. There may also be another whale joining into trail the flags and add their own to reduce rep further.

The spammers need to know that they cannot get away with this any more and we will not sit by and let our network be taken advantage of.

We do our best to make sure that no genuine accounts are in this list, even accounts in which the user uses pretty spammy behavior. @sadkitten is reserved at the moment for bot net scale spam. If you find that your account has been flagged in error, please get in touch in the comments or find me on steemit.chat

If you want to help find these accounts people get in touch with @spaminator or @sherlockholmes, not directly to me.

I will prepare some statistics of what the bot is doing when enough time has passed.

Anything else?

The time and expense I put into this I consider part of my witness activities, to help to protect the network by showing spammers they will not find these schemes profitable, and restore the reward pool for genuine content rewards. If you agree this is important please consider vote me as witnesses if you haven't already.

Vote for @personzzz (yes, three z's) for witness at https://steemit.com/~witnesses

Thanks to @transisto @patrice @pfunk and @sherlockholmes