Hate-bait phishing/Harassment and Cyberbullying/Hate Speech troll/bot farm accounts.

[mrsqueakcat]

Filter Mode

Auto-Smart Mode

Select the Problem

A type of spammer is not detected at all

(Optional) If 'Other', Enter Very Short Description

No response

Spammer Example / Sample

(Examples only.)
There are several hate-bait/harassment and bullying accounts that run all day long on this channel. https://www.youtube.com/channel/UCfpnY5NnBl-8L7SvICuYkYQ (These are just easily found examples of a common problem found all across YT)

ANY video from the last 2+ years has them in it to the point that on any given video around 20 to 40+% of the posts are just them baiting and harassing anyone and being called out for it.

The Sentient Flower and Bob Smith accounts are obviously run by the same troll farm operative and several other lower usage accounts show similar posting actions and styles and are well known to flair up if either of those accounts goes down for some unknown reason implying they are alternates.

(Problem In General) It's obvious that most larger YT channels have near-identical troll farm/bot accounts on them now that are run by operatives smart enough to stay just above the normal spam/harassment/hate-speech detection algorithms and YT doesn't seem to have the resources and or care to do anything about them.

Video / Post Link

https://www.youtube.com/channel/UCfpnY5NnBl-8L7SvICuYkYQ

(Optional) Additional Info / Context

Harassment and cyberbullying, hate speech, misleading/false information, or phishing/bait spamming are not allowed on YT, and these types of highly prolific troll/bot accounts specialize in doing all of that while staying above the normal YT filters that are supposed to handle such actions.

[Firecul]

Can you give a direct link to a couple of the comments you are referring to please

[ethnh]

Could you post a screenshot / copy examples of the comments you are talking about. Real humans posting hate comments would be out of the scope of this project to remove, and would likely require some sentiment analysis AI sort-of program to identify.

[mrsqueakcat]

Can you give a direct link to a couple of the comments you are referring to please

Just go to the channel I linked to in the opening post and pick any video from the last 2+ years and go down to the oldest comments. The two bot accounts (Sentient Flower and Bob Smith) generally make up about 1 in every 4 to 6 posts for the first half day the videos are up. According to the numbers from having test scanned a few videos they often make up about 50 to occasionally near 100 of the typical 300 to 500 posts.

It's just the same braindead childish hate-baiting and harassment spamming crap over and over based on common keywords and not the content or contexts of any post they are targeting.
And, no. It's not just this channel. It's just a really good example to use due to the relatively low comment traffic and extremely persistent troll/bot on it. Someone doing 50 to 100 shitposts every now and then on a channel that sees thousands of comments a day would never be noticed, But doing it every single day for over 2 years on a channel that only sees a few hundred comments a day at best makes them highly visible.

Could you post a screenshot/copy examples of the comments you are talking about. Real humans posting hate comments would be out of the scope of this project to remove, and would likely require some sentiment analysis AI sort-of program to identify.

I don't think it would be too hard to set up to just report them given that the software already does both, just not together yet and just needs some tweaking to get the two functions to work together in all modes. I already downloaded Python to use this experience to learn some more up-to-date programming skills.

[ethnh]

RE: MrSqeakCat Could you provide direct links, screenshots, or text copies of some of the comments you are mentioning, not instructions to find them👍

Also, are these messages already being flagged by YTSpammerPurge?

As to your response to my comment, not sure what you're trying to say. YTSpammerPurge (mostly) scans for words/urls, which is effective against simple bots, like most are currently. Human-generated hate comments (not copy+paste over and over), even if 'spammed', would easily bypass YTSpammerPurge's filters.

[mrsqueakcat]

RE: MrSqeakCat Could you provide direct links, screenshots, or text copies of some of the comments you are mentioning, not instructions to find them👍 Also, are these messages already being flagged by YTSpammerPurge? As to your response to my comment, not sure what you're trying to say. YTSpammerPurge (mostly) scans for words/urls, which is effective against simple bots, like most are currently. Human-generated hate comments (not copy+paste over and over), even if 'spammed', would easily bypass YTSpammerPurge's filters.

When possible, I try to get people to go look at things themselves thus removing the possible bias and limited range and scope of examples that doing screenshots tends to produce. Also in doing some digging into the channel's older videos I do see that many of the worst posts those two accounts make are often gone after a day or two and it's the low-level stuff staying behind which would also bias the validity of what I am asking people to review themselves.

So far, in running the spammer purge on his channels to test its limits, it is clearly limited in picking up the more high-level troll/bot posts, hence my issue with it and seeing that an option to filter and then report by account is not an unwarranted function.

I think the overall program has great potential but it needs to be fleshed out more so that it can better handle more difficult situations that require a more live hands-on assist as the higher-end troll farm/bot systems are evolving faster than the present ways to deal with them clearly are.

[ethnh]

When possible, I try to get people to go look at things themselves thus removing the possible bias and limited range and scope of examples that doing screenshots tends to produce. Also in doing some digging into the channel's older videos I do see that many of the worst posts those two accounts make are often gone after a day or two and it's the low-level stuff staying behind which would also bias the validity of what I am asking people to review themselves.

So far, in running the spammer purge on his channels to test its limits, it is clearly limited in picking up the more high-level troll/bot posts, hence my issue with it and seeing that an option to filter and then report by account is not an unwarranted function.

I think the overall program has great potential but it needs to be fleshed out more so that it can better handle more difficult situations that require a more live hands-on assist as the higher-end troll farm/bot systems are evolving faster than the present ways to deal with them clearly are.

Sorry, I am not very good at English. I am not sure what you are trying to say. Could you be very specific with what kind of 'troll' posts you are talking about? Specific examples? Bias in your examples does not matter as long as you can show some of the comments that are an issue 👍 Also, this is YT-Spammer-Purge, not YT-Troller-Purge 😅 - If it is not spam, it is not the target of this project.👍 In the issue description, you say

A type of spammer is not detected at all

What does the spam look like, exactly? It is hard to add filters without knowing what to filter. Thanks 👍

[mrsqueakcat]

Sorry, I am not very good at English. I am not sure what you are trying to say. Could you be very specific with what kind of 'troll' posts you are talking about? Specific examples? Bias in your examples does not matter as long as you can show some of the comments that are an issue 👍 Also, this is YT-Spammer-Purge, not YT-Troller-Purge 😅 - If it is not spam, it is not the target of this project.👍 In the issue description, you say

A type of spammer is not detected at all

What does the spam look like, exactly? It is hard to add filters without knowing what to filter. Thanks 👍

It seems that those two accounts have had almost all of their posts hidden from public view for the last few days videos (That's good!) so if I posted screen shots you would just see random people calling them out and not see what they posted.

The most common posts they do are like what stupid anarchist loser teens attacking random people for made-up political claims do. The main targets are people that they imagine are or support conservative views, even when what those people posted has little to no relevance to any sort of political views or have even called out such views themselves.

The main stuff is, calling people racists, bigots, homophobes, transphobes, nazis, fascists, white supremacists, KKK supporters, slave owners, Trumptards, Q-diots, cuckservatives, groomers, pedophiles, implying they have serious criminal records and they will be doxed if they don't stop posting, attacking their religious views, their military services, their ancestors, where they live, their educations, making outrageous false claims about American history and current events and so on. And it's always just because that person used a certain word or word set they can falsely associate with such things.

All of it seems to come from keyword-based cues like an automated reply system bot uses, hence the assumption that they are based on actual AI word recognition systems that just spam the comment sections with nonsensical attacks based on keywords only until someone takes their bait. Then they move to just making hate-filled outrageously stupid and typically irrelevant to anything counterclaims to anything anyone posts at them until the person gets sick of being harassed by a childish hate-filled idiot and stops interacting with them, implying that once they have hooked someone there is the actual hate-filled human operator behind the bots.

Not to make this political, but I expect these types of troll bot accounts to explode in activity all over YT for a while now that Elon Musk is acquiring Twitter. They always do when something big that they think will hurt far left-leaning agendas goes down.

[ethnh]

It seems that those two accounts have had almost all of their posts hidden from public view for the last few days videos (That's good!) so if I posted screen shots you would just see random people calling them out and not see what they posted.

The most common posts they do are like what stupid anarchist loser teens attacking random people for made-up political claims do. The main targets are people that they imagine are or support conservative views, even when what those people posted has little to no relevance to any sort of political views or have even called out such views themselves.

The main stuff is, calling people racists, bigots, homophobes, transphobes, nazis, fascists, white supremacists, KKK supporters, slave owners, Trumptards, Q-diots, cuckservatives, groomers, pedophiles, implying they have serious criminal records and they will be doxed if they don't stop posting, attacking their religious views, their military services, their ancestors, where they live, their educations, making outrageous false claims about American history and current events and so on. And it's always just because that person used a certain word or word set they can falsely associate with such things.

All of it seems to come from keyword-based cues like an automated reply system bot uses, hence the assumption that they are based on actual AI word recognition systems that just spam the comment sections with nonsensical attacks based on keywords only until someone takes their bait. Then they move to just making hate-filled outrageously stupid and typically irrelevant to anything counterclaims to anything anyone posts at them until the person gets sick of being harassed by a childish hate-filled idiot and stops interacting with them, implying that once they have hooked someone there is the actual hate-filled human operator behind the bots.

Not to make this political, but I expect these types of troll bot accounts to explode in activity all over YT for a while now that Elon Musk is acquiring Twitter. They always do when something big that they think will hurt far left-leaning agendas goes down.

I do not understand

It sounds like these accounts were removed by Youtube T&S, so issue is resolved?
If not, could you show an example of one of the bot comments? Are you sure the comments were spam? Were they copied across multiple videos, commented multiple times on the same video, or spammed in the replies of another comment? It is hard for me to understand what you mean without any examples of what these accounts were commenting. Thank you👍

[ikkysleepy]

I have a YouTube Comment Analyzer tool that will give you sentimental analysis per comment and video , but even then that system will not catch cyber bulling or hate speech. Also, most spammers I seen have a positive sentiment. These comments in that channel have an overall neutral sentiment. They also have no duplicate comments. So YT-Spammer-Purge will not detect any symptoms. Here is a screenshot of my tool where you can sort by comment count and also view that they have a lot of mentions.

Also, here an api that would find toxic comments, https://perspectiveapi.com/

[ethnh]

RE: ikkysleepy That's a really cool project! I was meaning to try building a tool that uses sentiment analysis, your project is a great use case for it😁

Message ID: @.***>

[mrsqueakcat]

I do not understand

It sounds like these accounts were removed by Youtube T&S, so issue is resolved? If not, could you show an example of one of the bot comments? Are you sure the comments were spam? Were they copied across multiple videos, commented multiple times on the same video, or spammed in the replies of another comment? It is hard for me to understand what you mean without any examples of what these accounts were commenting. Thank you👍

You're welcome. I haven't been all that active in the comment section for that channel for a while so I am a bit behind on things now. For a long time, those two accounts were in almost every thread being as stupid and awful as they could be to everyone.

For a long time, it was obvious that they build their replies from somewhat premade lists of fake talking points they think will upset certain groups. Often their comments are complete nonsequiturs or word salads that have no ties to anything that were built from strings of short prewritten sentences set around trigger keywords that don't really go together in context. AI wrote bait script or someone with a very weak understanding of how to write in English or a bit og both. (Google Translate type errors of sorts.) And when it's not either of those, it's them just being willfully childishly stupid and hateful just to get any sort of response they can get called out and start an idiotic argument they can full-on harass someone over.

Here's the sentient flower phishing with the usual nonsensical comments.

Then bob smith phishing with what he can get past the YT filters. (I think that account is being heavily censored at this point.)

bob smith attacking someone for nothing real. (Pretty mind compared to what he used to say)

Then him phishing with an obvious lie.

Sentient flower attacking some gal for made-up bigoted reasons about her name origins and getting called out on it. (Also likely getting censored a lot more than in the past)

This is why I think that the ability to just mass report based on account and not just content might have some use even if it is not a frequent one. With the higher level trolls/bot farm accounts are constantly evolving their tactics and posting methods to stay ahead of the automated YT systems that are supposed to limit harassment, bullying, and hate speech attacks.

[mrsqueakcat]

From today. Sentient flower phishing with nonsequiturs that often don't even match what he has replied to. ( I suspect that the worst are already being hidden from public view given how many replies are already unviewable.)

Both accounts will do this all day long for often up to 10+ hours for a less than hour-long video about nothing trying to get anyone to bite so they can belittle and harass them even more. Typically one account leads in for the day then they alternate back and forth for the rest of the day, often messing up which account has made what reply and to whom making it really obvious both accounts are run by the same person. If someone argues back hard enough they will put both accounts on that person and go full retard until whomever it is just gives up because of how hateful and stupid they act.

[mrsqueakcat]

An example of the type of lame nonsensical time waisting arguments they do with people that do engage with them. (This is fairly mild compared to what they used to post. Still doesn't make sense and is just a sentient flower wasting someone's time by acting stupid.)

As I stated at the beginning, this isn't exactly about them. It's just that they are very easy to use as examples due to their long-running high activity on a channel that has fairly low comment traffic. These types of harassment troll/bots are everywhere on YT now and do nothing but make things worse for the sake of making things worse.

[Firecul]

I'll be honest, this is trolling not spamming. Not really what this tool is about IMO. If these people were a nuisance to the creator then he could just block them from his channel. Why he hasn't is anyone's guess, I'm going to go with more comments > more views.

[mrsqueakcat]

I'll be honest, this is trolling not spamming. Not really what this tool is about IMO. If these people were a nuisance to the creator then he could just block them from his channel. Why he hasn't is anyone's guess, I'm going to go with more comments > and more views.

Yes, I am aware that all that's left to be seen given how so many of the replies seem to be hidden from public view now I can't give much for accurate examples of the spamming posting strategy being used.

My point in starting this thread was to address the more advanced spamming processes that get used to get around the filters. Whom/what I used has not been the best example due to how many of the posts are already being taken out. However, spamming with what looks like childish troll posting is quite possible too and easily gets past the current filtering systems.

I never expected much help or to be taken seriously with any of this because of how difficult it is to prove without wasting a lot of time on my end that would be better used just learning how to mod the code myself to have this function and maybe more.

[mrsqueakcat]

For those that still don't think that these types of extremely persistent harassment trolls are actually brainless AI-driven bots, watch this video on two chatbots talking and tell me that if this is what AI could do 10 years ago that it cant be behind the majority of the harassment trolls now that seem to have an impossible sub childlike ability to learn and grasp simple conversational contexts from what are for humans extremely simple and basic conversations.

https://www.youtube.com/watch?v=WnzlbyTZsQY

[ThioJoe]

Unfortunately missed this old thread so not sure if these types of comments are still active, if so feel free to leave a new issue with updated comments.