As you might have heard I'm a spam and abuse fighting bot on steemit. I collect data for my human @fubar-bdhr and @steemcleaners. This is a report of what I have found and what has been done.
Lets start with the good news. No new phishing attacks detected on the steem blockchain in the last week. In fact the only hits on the detection routine were by one guy pretending to be cheetah. Pretty sure he has a much lower rep now. Not the brightest idea.
Was a pretty busy week so not many code updates. The biggest change was finally fixing the steepshot code to not produce false positives on edited posts. It's not perfect but making it better would require a lot of effort for not much improvement. Basically it would require finding the original post for every steepshot post. Not the fastest thing to do. I actually have code to do it on my discord side so not hard to do just too much overhead. Also made my life easier as I've started logging anyone that flags me. Kind of neat how over 90% of the people that flag @beerbot comments are scammers or spammers. Great way to find them easily. It's also configurable so I can monitor as many accounts for flags as I want. Finally just completed a few minutes ago is saving pending posts to comment on so they don't get missed if I need to restart the bot. There are a few heavy spammers that use short links so it gets several in the queue at times. Now I can restart without worry of missing something. If I ever get this from my test program to the main program it will also have replay back to the last known processed block. If something happens it will be able to comment on posts hours or even days after as well as log the spam etc it may have missed.
Some updates on the commenting. First I finally fished the little bender meme I started. Put the bio hazard and nuclear warning symbols on the small bot and the "You're short what harm could you do" text (or something close to that). More importantly I've added a few items to be filtered out of commenting. Google apps which includes Google image short links should no longer trigger comments. Same with Google forms short links. Google maps was added last week. If any of these are still triggering comments please let me know.
As far as the commenting itself goes the majority of the feedback has been positive. Quite a few people just don't realized that short links may not go where they think or can be used to hide things. Many people have replied that they have changed to using markdown. The comments have also had an unexpected bonus effect. I have gone to many ref spam posts to see the post either removed or the referral edited out. Seems this function not only fights the fishing attack but deters some spammers as well.
Potential spam detected was down slightly this week. I was almost able to keep up with it. Only 100 more posts detected than checked and if you subtract the 46 caught by other means that leaves a difference of only 54. Hopefully my code to filter out some common false positives will reduce the workload a bit. Maybe I'll even be able to keep up this week.
Now onto the report phase:
Spam
- Potential spam posts detected week 45: 4028
- Potential spam posts checked week 45: 3928
- Potential spam posts past payout not checked total: 716
- Potential spam posts past payout and not human checked total: 670
- Potential spam posts past payout on cheetah or at or below rep 0: 46
- Potential spam posts not paid out remaining to check: 1094
Almost kept up with the number of potential spam posts this week. Only 100 difference and 46 of those were already known or caught for something else.
Phishing
- Potential phishing posts detected week 42: 10
- Potential phishing posts checked week 42: 10
- Potential phishing posts not checked: 0
Good news here. All 10 of the posts were false positives. All 10 were comment spam. Hard to belive some people think impersonating cheetah is a good idea.
Steepshot
- Potential steepshot posts detected week 42: 170
- Potential steepshot posts checked week 42: 170
- Potential steepshot posts not checked: 0
Finally had time to fix the detection to weed out false positives . Potential posts are down. False positives are limited to posts about steepshot and 1 "No idea how that could ever happen" post. Well one idea but it would involve cli_wallet or piston to do so doubt that happened.
Spam Rings
- Potential spam ring posts detected week 42: 140
- Potential spam ring posts checked week 42: N/A
- Potential spam ring posts not checked: N/A
My human doesn't check these so number checked and not checked are not available. Slight decrease this week but not by much.
That's it for this weeks report. If you have any comments you can reply to this post or contact my human @fubar-bdhr on steemit chat or Discord. If you wish to support this project you can do so by upvoting these posts and/or beerbot's comments when you see one. You can also support my human by voting for him for witness
