Coincidence or possible connections between accounts?!

[jahead]

As part of another project, I created a "Multi-Account-Scanner" to find different ways to spot duplicate accounts.

I had the idea to compare sleep/inactive times and posting histories to find patterns between users using my own algorithm. I don't know if this is enough on its own an wanted to share with you guys.

Ok i know, of course it could be a coincidence, but if you look at the screenshots (which I blurred for now), you can see several strange patterns. Atm this screenshots show members from just two campaigns that I compared to each other. This means everyone shown is a participant in one of two campaigns:

     
   

Registration Dates: The accounts were created very close to each other.
Activity & Idle Times: Their posting times and breaks match in a suspicious way.
Posting Intervals: The time between posts suggests maybe that one person has multiple browser tabs open and switches between accounts, causing short delays of just a few minutes.

Funfact: When looking closely at some of the posts, a lot of them seem to be AI-generated (i think AI-Detector will not say so) because they contain nonsense or generic filler content.
Because of this, I could bet that there is a bot in use, using AI to control these accounts automatically or letting a human post automatically at the click of a button.
I think of this, bc a frequent pattern stands out - an account really often quotes a post and replies with a short 3-line response. I noticed this across a few accounts, which suggests an automated process where AI generates replies to quoted content. YES, this is just my theory, but it seems possible.

What is your opinion on this? Could so many coincidences really happen at once? Or are these enough clues to investigate further and reveal their names? What shoul I do?

This are just examples and I can say that if I add more campaigns, I promise I will find many more accounts.  


edit: Just to emphasize the "coincidence," in case it was missed:

Here are multiple pairs of separate accounts (all in sig-campaigns) that both of them posted for example at least 17 times within a single month (in 30 days) - on the same day and during the exact same hour, their posts were only 3 to 30 minutes apart. And this applies to all the pairs here.  


Edit²:

Glad I left the audience speechless.  

I know these are just strong clues, not solid proof, and we are missing the final step.

I will keep this thread open to track, improve, and document the progress of my "algorithm and scanner".

Feel free to share your ideas, feedback, and questions!  

[Free Market Capitalist]

bump!
Glad I left the audience speechless.  

You are breaking the rules with that bump because it hasn't even been 24 hours yet since your first post. I've reported it to be merged.

Instead of wondering if you left the audience speechless, you might want to think about what you did wrong that caused no one to respond. You'd do much better.

[jahead]

Thanks for pointing out the "bump" rule, I honestly missed that in the heat of the moment.  

Trust me, I’m thinking about why everyone is so quiet.   I’d love to get your feedback and suggestions. It sounds like you might be able to help me improve myself, my scanner and the methodic, so please, fire away and give me some hints please!  

edit: my PMs are open as well ofc.  

[examplens]

Trust me, I’m thinking about why everyone is so quiet.   I’d love to get your feedback and suggestions. It sounds like you might be able to help me improve myself, my scanner and the methodic, so please, fire away and give me some hints please!  

edit: my PMs are open as well ofc.  

You presented some theory of your own, which might make sense, but you hide the names of the accounts. How do you expect anyone to give their opinion or analysis if they can't check it themselves? You are still trying to leave the discussion in PM.
Why are you afraid to write which accounts these are?

[hugeblack]

If there is such evidence, post it under the name “possible alt,” and if more likely connections are found, for example, “link between addresses,” same social media ID, same writing style.

edit: my PMs are open as well ofc.  

This is a bad way to promote your service. If the code is open source, move the topic to Project Development.

[nutildah]

Its an interesting approach but you have to be careful not to make any accusations based on this alone. Sometimes what looks like patterns is just pure coincidence, and its easy to find oneself connecting dots that just aren't there.

Once you tie it to something more firm like blockchain links, then it will become more useful to demonstrate an association.

[YOSHIE]

As part of another project, I created a "Multi-Account-Scanner" to find different ways to spot duplicate accounts.

This is the Reputation section/before you accuse or use "Multi-Account" you first prepare evidence that is accurate and reasonable, when seen by other members.

You can take an example of the format here:
https://bitcointalk.org/index.php?topic=5394559.0

Remember, when you have created the topic of Multi-Account accusations, there is no word "Coincidence or possibility" you can go straight to the point, include valid evidence.

I saw that the picture you posted said Rainbet.com, do you mean Multi-Account in the Rainbet.com campaign or Multi-Account at the casino, I'm still confused about your topic, just getting straight to the point is better.
Or post the names of accounts that you are suspicious of, right?

[AakZaki]

What if you reveal these names without accusing them first, such as grouping names that according to your data are suspicious.
If possible, I will re-investigate and find out the truth, whether they are indeed connected accounts or not.

Because if you don't mention those names, all of us here can't know. It could even be that the tool will not function as it should.

[Awaklara]

OP, I have a suggestion if you feel uncomfortable publishing those accounts. You could try your tool with accounts that have already been published before. Just to see the data and how your tool presents it.

lovesmayfamilis has previously published cases related to similar patterns on several accounts. You can try: https://bitcointalk.org/index.php?topic=5394559.msg66018613#msg66018613

The idea is actually interesting, it could be used as a tool to assist investigations.

[jahead]

At first, thanks a lot for all your input and feedback.
It really helps me to improve myself and this project on many levels. As you can tell, I still need help to figure things out and feel quite unsure about this topic.

[...] You are still trying to leave the discussion in PM. [...]

Sorry - bc of my "open PMs",
I thought there might be something, somebody want to share about my methods, but that could help farmers/multi-accs prevent this - which is why I thought they shouldnt maybe be public.
Stupid of mine, nevermind.

[...] Why are you afraid to write which accounts these are?

I hide the account names until now, because I didnt want to make false accusations.
Im unsure if my evidence is strong enough to post the profiles names. Based on some comments of you, it seems maybe it is - maybe its not...  

Its an interesting approach but you have to be careful not to make any accusations based on this alone. Sometimes what looks like patterns is just pure coincidence, and its easy to find oneself connecting dots that just aren't there.

Once you tie it to something more firm like blockchain links, then it will become more useful to demonstrate an association.

I hear you. For a short time, it could be just a coincidence that two accounts always post so close together... I could understand that, still sus. But if this happens over a very long period, I can no longer believe it is a coincidence for two completely independent accounts (especially since these two were registered so close in time together). This means I need to expand the timeframe of the posts I search for, to make my argument and findings stronger.

This is the Reputation section/before you accuse or use "Multi-Account" you first prepare evidence that is accurate and reasonable, when seen by other members.

You can take an example of the format here:
https://bitcointalk.org/index.php?topic=5394559.0 [...]

Awesome, thanks! I will look at the thread and the format in detail as an example.

[...] I saw that the picture you posted said Rainbet.com, do you mean Multi-Account in the Rainbet.com campaign or Multi-Account at the casino, I'm still confused about your topic, just getting straight to the point is better. [...]

As mentioned in the opening post, I have so far only compared the profiles of the "Rainbet"- and "Rollbit"-signature-campaigns, and I already found these suspicious patterns between users there.
In the screenshots, the text behind the blacked-out names always shows which signature campaign the user belongs to.

What if you reveal these names without accusing them first, such as grouping names that according to your data are suspicious.
If possible, I will re-investigate and find out the truth, whether they are indeed connected accounts or not.

Because if you don't mention those names, all of us here can't know. It could even be that the tool will not function as it should.

You are completely right on your points, thank you for that. As I said, I will first update my scanner and adjust a few settings to minimize "coincidences."
In the next days, I will present the results here with the respective profiles, so they can be double-checked then to see if the accounts are actually connected.

OP, I have a suggestion if you feel uncomfortable publishing those accounts. You could try your tool with accounts that have already been published before. Just to see the data and how your tool presents it.
[...]
The idea is actually interesting, it could be used as a tool to assist investigations.

Testing the scanner against the existing results is a really great idea, thank you for that!
I will do exactly that and see if it gives me even more starting points to work with. Very nice!