💤 Inactive Cozy.tv - Spiraling unsustainable honeypot.

[Squirrelcop]

Could the weird spikes maybe be explained by the fact that people log off the streams to go to bed before others who just woke up/just went through their morning routine then jump in?

I can also imagine that a part of the viewership could be from UK and Australia or other parts of the world (Germany also has a few Alt Right schlemils who love to get their material from their American counterparts).

Still suspicious, tho.

Also awesome job on the OP @AltisticRight . Its a great start off point for people like me who are new to the whole cozy.tv thing.

 

[HatredOfScissors]

You should mention the live viewer multiplier built into the backend that Simon talked about @AltisticRight.
None of their view counts are real and it's really obvious with Ralph when he has between 50-200 viewers on YouTube and Odysee but somehow over 1k on cozy.

Anyone with a brain knew the viewers were fake and simon just confirmed it. Fuentes supposedly gets north of 6000 every night but his superchat bot loops back around to the first one and theres the clip of him looking at his superchats and asking "whats the date again?"

As for the warthog, his viewership mysteriously climbed to well over a thousand for that margartita sunrise shit, the only livestream show classified as a certified tranquilizer, for 8 hours a day. I imagine they gave him access to his own view multiplier and hes been power tripping making it higher and higher with no regard to what makes sense or is believable.

 

[Snigger]

Anyone with a brain knew the viewers were fake and simon just confirmed it. Fuentes supposedly gets north of 6000 every night but his superchat bot loops back around to the first one and theres the clip of him looking at his superchats and asking "whats the date again?"

As for the warthog, his viewership mysteriously climbed to well over a thousand for that margartita sunrise shit, the only livestream show classified as a certified tranquilizer, for 8 hours a day. I imagine they gave him access to his own view multiplier and hes been power tripping making it higher and higher with no regard to what makes sense or is believable.

We're also suspicious of the follower count, especially for people like Ralph. There's a general, consistent upward trend in followers across the site yet viewership is relatively stable.

 

[Gamercat]

how long before cozy tv goes full stream.me and just dies?

 

[Snigger]

how long before cozy tv goes full stream.me and just dies?

It won't last a year at this rate, I'm amazed they haven't been attacked like the farms has yet, they're a much more prominent target.

 

[temp o'rary]

OK, so here's how you detect a simple fixed multiplier scheme of the kind thought to have been described in the KC stream:

Have a look over the ranges for which you have plenty of data. Say for example the 7000-8000 range (I haven't looked to see if this one is useful, I just have it in my mind as Nick's approx viewer count).

If the multiplier was exactly 2, then not only would we have "just evens" (easy to detect by eye), but we'd also only have 500 possible values in that range. 7001 wouldn't appear because it corresponds to a real viewcount of 3500.5, and the same for 7003, 7005, ... etc.

If the multiplier was odd, say 3, as someone suggested, then it's harder to eyeball but now we only have ~333 possible values (multiples of 3). They'd be 7002, 7005, etc...

"But what if they're being sneaky and multiplying by 2.12364234 then rounding?" (what people think Simon was alluding to, to avoid "no evens")

Same principle. The numbers 1-1000 will be mapped to something like 2-2123 but there will still only be 1000 "slots" used in this fake range. And in the 7000-8000 range we'll still end up with (8000-7000)/2.12364234 ~= 471 possible fake viewcount values.

So, look at the number of unique values taken by the fake viewcount, preferably in ranges where you have plenty of data. (I asked how much data because there's a simple workaround which defeats this method going forward). Plotting those unique values will reveal somewhat regular "gaps" (but not completely regular if rounding and fractional multipliers are being used).

To get better results, filter out the consecutive identical viewcounts where they haven't updated the count yet. With those removed, spikes and gaps in the frequency plot will be more meaningful (as the consecutive counts will produce spikes and gaps even if they weren't fixing the numbers - some viewcounts will get multiple hits simply because you've collected the un-updated value multiple times)

It might be complicated if they were changing the multiplier value over time (you'd have to analyze separate time periods, then you've got less data), or if they anticipated this kind of analysis and incorporated a simple fix (I won't give them any hints here), or they're using a different scheme entirely. But this will detect any fixed multiplier >1, given sufficient data.

tl;dr: "proof" will look something like this: "We've collected 2356 independent viewcounts in the 7000-8000 range, but they only appear as 480 unique viewcount values. Without a multiplier, you'd expect them to fill most of the range!"

(even if you have less data for the range, say 400, you'd still expect that to end up as roughly 400 values, and certainly not 200. but if you have fewer datapoints in a range than there are slots, it's possible for them all to take unique slots and look "innocent", even if the multiplier is in effect. having more data than slots eliminates this possibility)

 

[Gamercat]

OK, so here's how you detect a simple fixed multiplier scheme of the kind thought to have been described in the KC stream:

Have a look over the ranges for which you have plenty of data. Say for example the 7000-8000 range (I haven't looked to see if this one is useful, I just have it in my mind as Nick's approx viewer count).

If the multiplier was exactly 2, then not only would we have "just evens" (easy to detect by eye), but we'd also only have 500 possible values in that range. 7001 wouldn't appear because it corresponds to a real viewcount of 3500.5, and the same for 7003, 7005, ... etc.

If the multiplier was odd, say 3, as someone suggested, then it's harder to eyeball but now we only have ~333 possible values (multiples of 3). They'd be 7002, 7005, etc...

"But what if they're being sneaky and multiplying by 2.12364234 then rounding?" (what people think Simon was alluding to, to avoid "no evens")

Same principle. The numbers 1-1000 will be mapped to something like 2-2123 but there will still only be 1000 "slots" used in this fake range. And in the 7000-8000 range we'll still end up with (8000-7000)/2.12364234 ~= 471 possible fake viewcount values.

So, look at the number of unique values taken by the fake viewcount, preferably in ranges where you have plenty of data. (I asked how much data because there's a simple workaround which defeats this method going forward). Plotting those unique values will reveal somewhat regular "gaps" (but not completely regular if rounding and fractional multipliers are being used).

To get better results, filter out the consecutive identical viewcounts where they haven't updated the count yet. With those removed, spikes and gaps in the frequency plot will be more meaningful (as the consecutive counts will produce spikes and gaps even if they weren't fixing the numbers - some viewcounts will get multiple hits simply because you've collected the un-updated value multiple times)

It might be complicated if they were changing the multiplier value over time (you'd have to analyze separate time periods, then you've got less data), or if they anticipated this kind of analysis and incorporated a simple fix (I won't give them any hints here), or they're using a different scheme entirely. But this will detect any fixed multiplier >1, given sufficient data.

tl;dr: "proof" will look something like this: "We've collected 2356 independent viewcounts in the 7000-8000 range, but they only appear as 480 unique viewcount values. Without a multiplier, you'd expect them to fill most of the range!"

(even if you have less data for the range, say 400, you'd still expect that to end up as roughly 400 values, and certainly not 200. but if you have fewer datapoints in a range than there are slots, it's possible for them all to take unique slots and look "innocent", even if the multiplier is in effect. having more data than slots eliminates this possibility)

big fat pepe frog kids destroyed by facts and logic!

 

[Snigger]

OK, so here's how you detect a simple fixed multiplier scheme of the kind thought to have been described in the KC stream:

Have a look over the ranges for which you have plenty of data. Say for example the 7000-8000 range (I haven't looked to see if this one is useful, I just have it in my mind as Nick's approx viewer count).

If the multiplier was exactly 2, then not only would we have "just evens" (easy to detect by eye), but we'd also only have 500 possible values in that range. 7001 wouldn't appear because it corresponds to a real viewcount of 3500.5, and the same for 7003, 7005, ... etc.

If the multiplier was odd, say 3, as someone suggested, then it's harder to eyeball but now we only have ~333 possible values (multiples of 3). They'd be 7002, 7005, etc...

"But what if they're being sneaky and multiplying by 2.12364234 then rounding?" (what people think Simon was alluding to, to avoid "no evens")

Same principle. The numbers 1-1000 will be mapped to something like 2-2123 but there will still only be 1000 "slots" used in this fake range. And in the 7000-8000 range we'll still end up with (8000-7000)/2.12364234 ~= 471 possible fake viewcount values.

So, look at the number of unique values taken by the fake viewcount, preferably in ranges where you have plenty of data. (I asked how much data because there's a simple workaround which defeats this method going forward). Plotting those unique values will reveal somewhat regular "gaps" (but not completely regular if rounding and fractional multipliers are being used).

To get better results, filter out the consecutive identical viewcounts where they haven't updated the count yet. With those removed, spikes and gaps in the frequency plot will be more meaningful (as the consecutive counts will produce spikes and gaps even if they weren't fixing the numbers - some viewcounts will get multiple hits simply because you've collected the un-updated value multiple times)

It might be complicated if they were changing the multiplier value over time (you'd have to analyze separate time periods, then you've got less data), or if they anticipated this kind of analysis and incorporated a simple fix (I won't give them any hints here), or they're using a different scheme entirely. But this will detect any fixed multiplier >1, given sufficient data.

tl;dr: "proof" will look something like this: "We've collected 2356 independent viewcounts in the 7000-8000 range, but they only appear as 480 unique viewcount values. Without a multiplier, you'd expect them to fill most of the range!"

(even if you have less data for the range, say 400, you'd still expect that to end up as roughly 400 values, and certainly not 200. but if you have fewer datapoints in a range than there are slots, it's possible for them all to take unique slots and look "innocent", even if the multiplier is in effect. having more data than slots eliminates this possibility)

I'm just focusing on collection right now, I'm currently taking snapshots of their API page every ~3 minutes. I've got about 12MB of data already.

 

[AltisticRight]

mabey its the autism but simons answer was basically that they multiply by a long string of decimals so that it will never be only odd or even. which is a autistic way to say they use fractions

Yeah, that's kind of what he meant, saying there's ways to it and programmers aren't stupid. I don't think he's autistic, he's obviously a smart guy but this code stuff doesn't seem like his cup of tea, neither is mine, I'll just play with the numbers. Problem is, this multiplier thing honestly won't work well, you'll end up with a finite pool of values and people can easily find out the multiplier, it's just a matter of a massive dataset, something we actually have now. Maybe I'll just run it, but could they be that fucking stupid?

So, look at the number of unique values taken by the fake viewcount, preferably in ranges where you have plenty of data.

Thanks for this comment. They could just use a sneakier method by generating some stupid fraction each day and dynamically finetuning it by random time intervals. I don't think it's worthwhile but they could be desperate.

Could the weird spikes maybe be explained by the fact that people log off the streams to go to bed before others who just woke up/just went through their morning routine then jump in?

It's genuine and happens once in a blue moon. Ethan Ralph's spikes were his Jim vs. Nick restream and some worthless AF interview.

 

[temp o'rary]

You can do the integer multiplier thing undetectably: you multiply by 2, then randomly add a 1 half of the time. Or you can multiply by 3, and add one of [0,1,2] each with equal probability. And so on.

The fact that he mentioned the "trick" with the number close to 2 seems a bit specific, as if it's what they actually went with. It's bad, but then it took them months to get viewer counts showing, and they probably weren't paranoid enough to expect frequency analysis.

 

[HatredOfScissors]

Im not sure if this has been done or proposed before, but a way to get an estimate of their actual live viewers would be to look at the ratio of unique chatters to live viewers. On cozy this should be small enough to tally up by hand. Then compare it to legitimate platforms like youtube and odysee, and other livestreamers.

 

[NetVet]

Don't want to spam this discussion with my dumb remarks but maybe the fakery could be backed up if we look at the engagement of his Telegram?

Average sub to Nick's Telegram is his devotee and a sycophant so one would expect they're all over his posts and updates?

45K members
11K saw this
roughly 500 of them interacted?

Not that this is proof of anything, just kinda seems that if the Telegram psychos aren't engaged, who the fuck else would go to Cozy.tv? I'd bet 99% of Cozy viewers are following him on Telegram, it's just a echo-litterbox (funny amirite).

 

[ElAbominacion]

Yeah, that's kind of what he meant, saying there's ways to it and programmers aren't stupid. I don't think he's autistic, he's obviously a smart guy but this code stuff doesn't seem like his cup of tea, neither is mine, I'll just play with the numbers. Problem is, this multiplier thing honestly won't work well, you'll end up with a finite pool of values and people can easily find out the multiplier, it's just a matter of a massive dataset, something we actually have now. Maybe I'll just run it, but could they be that fucking stupid?


Thanks for this comment. They could just use a sneakier method by generating some stupid fraction each day and dynamically finetuning it by random time intervals. I don't think it's worthwhile but they could be desperate.


It's genuine and happens once in a blue moon. Ethan Ralph's spikes were his Jim vs. Nick restream and some worthless AF interview.

If these people were smart they'd have used any of the applications of 1 plus random() that exist in web development languages and add a conditional to only apply it when it's positive AND above a certain threshold.

 

[Total Annihilation]

I find it quite funny how the website uses Telegram as a login method; aka the same messaging platform with a backdoor built in.
Even if cozy doesn’t already have a backdoor built in; given that it uses Telegram as a login method, it wouldn’t surprise me that it’d be easy to get into.

 

[blacktrollface]

Not sure if this is something that should be shared here publicly, but i'm not a computer guy so hopefully one of you guys knows what to do

This is the cozytv studio dashboard login page: https://cozy.tv/studio/

When you put in an invalid password and username combo, it's not rate limited, which means you can do it over and over and over again as much as you want without anything happening.

From what I read in a chat i'm in, this is very insecure, and could actually be scripted to where you could do it so many times it would crash the page or cause some type of damage. It obviously won't crash cozy itself, but the streamers probably wouldn't be able to stream if you crashed their dashboard.

Have fun.

 

[SploogeInspectorGeneral]

Not sure if this is something that should be shared here publicly, but i'm not a computer guy so hopefully one of you guys knows what to do

This is the cozytv studio dashboard login page: https://cozy.tv/studio/

When you put in an invalid password and username combo, it's not rate limited, which means you can do it over and over and over again as much as you want without anything happening.

From what I read in a chat i'm in, this is very insecure, and could actually be scripted to where you could do it so many times it would crash the page or cause some type of damage. It obviously won't crash cozy itself, but the streamers probably wouldn't be able to stream if you crashed their dashboard.

Have fun.

Is it possible to do this without a script? If a large group mass submits invalid login info over and over, would that do it?

Been smashing login button for the last 5 minutes but don't want to waste my time. Now that this info is out there though, surely there will be cozy developers watching so anything you want to do, better get to it before they patch.

 

[blacktrollface]

If enough people tried it at the same time it's possible, but it's way more efficient to run a script since it can do it so much faster, and will just continue doing it 24/7

I have no idea how to do that, and you'd probably have to have more than one computer running the script to actually fuck with the site.

Not many people watch this thread, hopefully people will take notice before cozy devs try to do something.

 

[AltisticRight]

Feeling cosy there in your autistic grifty incubator, Nicky?
@Perspicacity Have fun with this.

First, we caught a certain piggy with his hooves deep into the jar of cookies. How did we know that? Well, AFF cookies are baked with a special ingredient that reacts to black light, let's just say it's a natural ingredient high in protein.
2 days of streaming, countless hours a day, Ethan Ralph gained less than 100 followers. This equates to around 40 followers a day! Totally not a stagnant shithole.

Let's examine the rage pig's viewers.

Does this look real to anyone? The viewership is constant, takes a dive than skyrockets to over 200, and then dives down to several hundred. This isn't that Luna cryptoshekel.
There is definitely viewership manipulation. Giving them all the benefits of the doubt, the viewership indicator is not truthful and does not function properly.

Secondly, let's look at a 4'8'' soy drenched goblin.

A whopping 8-follower-gain in the span of 7 days. How does one use this platform and not want to kill themselves?

With a similar subscriber count to Ethan Ralph, his engagement is way lower, sad. The same completely bizarre trend is seen.

Finally, the catboy fuhrer himself.

Same story, less than 100 followers gained over the span of these days.

Yeah right. Does this look like natural viewership to anyone?

In fact, this shitcoin pump and dump style nosedive was recorded by someone.

In conclusion, there's something very very fishy about these numbers. They are completely fabricated, they have no resemblance to anything natural. Put forward any theory you like, here's mine.
I believe they fix the live viewer count to a certain number for each streamer. Once people pour in, something weird in their code causes the number to get recalculated at that instance. Then, the viewership slowly climbs at a steady rate.
Anyway, Nick's follower gain is approximately the same as Ralph's, yet he supposedly has 8000 watching. Ralph and Beardson's follower count is similar, yet Ralph gets far more viewers.

 

[gaystoner]

Let's be honest...Who's ready for PPP to triple the audience of the entire site of Cozy.tv even with nick streaming?

hysterical. faggots.

 

[Sam Losco]

How hard would it have been to code it so it's not so obviously fake?
When the streamer goes live, make the live viewer count, say, 10-20% of that streamers followers/subscribers. Then, increase that number up by .5-1% every 2 minutes until it maxes at 66-85% of total followers/subscribers. Periodically randomly increase or decrease by .2-.5% every 2 minutes.
There, that seems like it would be way more natural looking than this weird bullshit they are doing.