Twitter automation test (300 accounts, 6 control groups)

Report
6/28 ~ 7/1


This is a screenshot of datadog logs.
If you see below, I have 1000 actions sets done successfully.

스크린샷 2023-07-01 오후 4.50.03.png


Round 1 Test Results (300/300)
Success
Round 2 Test Results (300/300)
13 accounts banned
Round 3 Test Results (300/300)
16 accounts banned
Round 4 Test Results (100/300)
Currently running

Total 29 accounts banned
set 4 count = 4
set 5 count = 25

Unfortunately, I have 16 more accounts blocked.
All the bans happened in testset 4, 5.
So I must be wrong in the last report.
I am just assuming it happens because of mass following.
I cannot really say for sure.

1688198208294.png


1688198460043.png
 
Last edited:
Very interesting thread. I have 3 questions for you hope you will answer.
1. Did you try out antidetect browser for automation?
2. Did twitter or other detect you are a bot through UDP traffic, if you are only using TCP (sock5, proxy)...
3. How to manage all account logs with Datadog. Any guide for this i can find over internet?
Thank you.
 
Very interesting thread. I have 3 questions for you hope you will answer.
1. Did you try out antidetect browser for automation?
2. Did twitter or other detect you are a bot through UDP traffic, if you are only using TCP (sock5, proxy)...
3. How to manage all account logs with Datadog. Any guide for this i can find over internet?
Thank you.

1. Yeah I have. But I have ditched that because of the following reasons.
https://www.blackhatworld.com/seo/t...t-4-factors-32-accounts.1461652/post-15939602
2. No. I had not had any problems with the proxy being detected. I am only using http proxies.
A website can detect we are using a proxy. But using a proxy itself is not a red flag as normal people can do too.
I also doubt that the web, or browsers leak such information about the users computer networks.
That would be a serious flaw in web security.

3. Finally someone who appreciates logging and data analytics!!
We definitely need more people who code in this community.
This is for python. but I am sure you can google and find datadog for your language.
https://github.com/DataDog/datadog-api-client-pythonI advise you not to use agents (datadog program that checks for memory, cpu usage, makes logging easier if installed)
They cost 15$ per machine.
 
do you know the reason cause funcaptcha verify? I have the problem when i mass like the same tweet
 
I do not think that arguments need to be emotional nor personal.
It is just a way to find out what is true or better.

1. I do not know when you have tested this, but in my previous test and my current test, outlook is passing with no problemo.
previous test : https://www.blackhatworld.com/seo/twitter-automization-test-4-factors-32-accounts.1461652/

2. Blocked numbers is agreed!

3, 4. 48k actions with 300 accounts. I am not being detected.
Well the prevous report had 16 banned. And I doubt this was because of the browser.
If I was detected by browser, I should have bans all over the place. (not in 2 min timeframe, check my previous report)
View attachment 266171

5. 6. I don't know if this is true or not.
Thats why I have test groups separated by pyautogui, and selenium clicks.
But thanks for the link. Checked it out, seems great.
I am using python, so I would need to change this to a shell tool in order to implement it.
These are only few tips on how you can improve!

Browser leaks are so common these days, every week a new update is implemented.

I have attached a image below for event detection!
Screenshot 2023-07-01 at 4.29.17 PM.png


although it's great to see you testing this way, Hope to see things work out for you as expect.
 
These are only few tips on how you can improve!

Browser leaks are so common these days, every week a new update is implemented.

I have attached a image below for event detection!View attachment 266236

although it's great to see you testing this way, Hope to see things work out for you as expect.
nice find :)
will have to check this out.
Don't you think creepjs gives way too many false positives, though? Well, not exactly false positives but its almost impossible to get a decent score on it unless you're using vanilla chrome. Even vanilla firefox gets a poor rating.
 
These are only few tips on how you can improve!

Browser leaks are so common these days, every week a new update is implemented.

I have attached a image below for event detection!
although it's great to see you testing this way, Hope to see things work out for you as expect.
holy shit, you found the source js code of twitter that detects events mouse cilck detection!!!!
I frekin love this place.

You have 100% confirmed.
Twitter detects mouse movement and clicks.

Kudos to you mate!

mouse move.PNG
 
nice find :)
will have to check this out.
Don't you think creepjs gives way too many false positives, though? Well, not exactly false positives but its almost impossible to get a decent score on it unless you're using vanilla chrome. Even vanilla firefox gets a poor rating.
The thing with creepjs is they know what Fingerprint data is missing or doesn't belong, Suppose you are using Windows Fingerprint and you have Fonts of Mac, it will minus points for it.

Now the thing with Fingerprints is the data should belong to the right place, not in some random arrays.

This mistake is very common, even if a guy develops a anti detect browser he still thinks that the random data will do the work, but it won't solve any problem it will increase the detection rate!

Why? Many companies have already started implementing AI for detections, they know what a good Fingerprint looks like, of-course with so much data everything is possible.

but Good fingerprint also won't solve this, if you have noticed creepjs have a thing where it increases trust score over the time if you are a good user, even if you have a good fingerprint doesn't always means good scores, a good user with history is also an important part for detection, a good web history aswell.
 
do you know the reason cause funcaptcha verify? I have the problem when i mass like the same tweet

At first I have used phones but still it blocks you after mass traffic.
I think Twitter detects the ip cidr block (think of it as a range of ip), internet service provider, phone fingerprint, webgl fingerprint and all that stuff.
And they makes this as a untrustworthy combination.
If you send too much traffic in these combinations or properties, twitter asks for funcaptcha.

But in your case, I think thats the bot detector AI, not the register captcha,
The action of mass liking the same tweet with the same fingerprint leaking can be flagged. (This is a suspection. have not tested)
 
The thing with creepjs is they know what Fingerprint data is missing or doesn't belong, Suppose you are using Windows Fingerprint and you have Fonts of Mac, it will minus points for it.

Now the thing with Fingerprints is the data should belong to the right place, not in some random arrays.

This mistake is very common, even if a guy develops a anti detect browser he still thinks that the random data will do the work, but it won't solve any problem it will increase the detection rate!

Why? Many companies have already started implementing AI for detections, they know what a good Fingerprint looks like, of-course with so much data everything is possible.

but Good fingerprint also won't solve this, if you have noticed creepjs have a thing where it increases trust score over the time if you are a good user, even if you have a good fingerprint doesn't always means good scores, a good user with history is also an important part for detection, a good web history aswell.
I've been trying to make a custom cloaked browser, have managed to spoof most things, but I keep failing on the "serviceworkers useragent" thing. It seems there isn't any way around service workers other than disabling them. Thoughts?
 
I've been trying to make a custom cloaked browser, have managed to spoof most things, but I keep failing on the "serviceworkers useragent" thing. It seems there isn't any way around service workers other than disabling them. Thoughts?
There is a way around workers, it's lengthy, it's the most common leak, i think you should save some headache and time by disabling them majority of the detections don't work around workers, you won't see any such detections out in the wild.
 
7/2 ~ 7/3
Test report


Round 1 Test Results (300/300)

Success
Round 2 Test Results (300/300)
13 accounts banned
Round 3 Test Results (300/300)
3 accounts banned
Round 4 Test Results (300/300)
Success
Round 5 Test Results (0/300)
Currently running

No additional bans found in round 4.

Total 1.2k successful actions
스크린샷 2023-07-03 오후 10.45.36.png




Sorry to say, I have misinformed you guys about the ban count.
Unfortunately, I have 16 more accounts blocked.
The ban logs count for every test set is the total count of banned accounts.
I thought it was the added amount.
So the real ban count would be...

Round 2
13 accounts banned
11 (Set 5) + 2 (Set 4)

Round 3
3 accounts banned
3 (Set 5)


As you can see the account seq of the logs is duplicated.
So the previous accounts logged as banned will be logged as banned again for each set.

스크린샷 2023-07-03 오후 10.58.26.png



The tests are taking soo slow.
I am thinking of expanding another machine with proxy.
We are 1/7 done until the final test.
This is gonna take a long time...
 

Attachments

  • 스크린샷 2023-07-03 오후 10.46.50.png
    스크린샷 2023-07-03 오후 10.46.50.png
    75.5 KB · Views: 158
7/4 ~ 7/9
Test report


Round 1 Test Results (300/300)

Success
Round 2 Test Results (300/300)
13 accounts banned
Round 3 Test Results (300/300)
3 accounts banned
Round 4 Test Results (300/300)
Success
Round 5 Test Results (180/300)
Currently running


Banned Accounts

Round 2

13 accounts banned
11 (Set 5) + 2 (Set 4)

Round 3
3 accounts banned
3 (Set 5)


Issues
I had some problems while running my machines.
In the 6/20 issue in page 2 we had memory issues while running chrome.
Apparently, when running undetected chromedriver we can have storage issues too. XD

스크린샷 2023-07-09 오후 12.02.08.png

스크린샷 2023-07-09 오전 9.02.15.png


I did eventually find out where all these GBs of storage is being used.
It seems that undetected chromedriver downloads a chromedriver file in every run.
Code:
/tmp/snap-private-tmp/snap.chromium

스크린샷 2023-07-09 오전 9.07.12.png


To remove all temporary chromedrivers
Code:
rm -r /tmp/snap-private-tmp/snap.chromium/*

I did remove all the temp files manually for now.
But I will program this to run automatically.

스크린샷 2023-07-09 오전 9.14.20.png



Improvements
I was running with odroid n2l, but changed to beelink eventually.
I was suggested to reboot and clear the cache of chromium browser when I had the memory issues.
But after I have changed this, it took too much more time when rebooting and slowed down the process.

I did order another beelink machine and another lte sim.
So I will be run with 3 machines concurrently soon.
Wish this can speed up the process.
It does come from aliexpress so, it takes like 12 days to arrive here. Xd

Have a nice chill day.
 
7/4 ~ 7/9
Test report


Round 1 Test Results (300/300)

Success
Round 2 Test Results (300/300)
13 accounts banned
Round 3 Test Results (300/300)
3 accounts banned
Round 4 Test Results (300/300)
Success
Round 5 Test Results (180/300)
Currently running


Banned Accounts

Round 2

13 accounts banned
11 (Set 5) + 2 (Set 4)

Round 3
3 accounts banned
3 (Set 5)


Issues
I had some problems while running my machines.
In the 6/20 issue in page 2 we had memory issues while running chrome.
Apparently, when running undetected chromedriver we can have storage issues too. XD

View attachment 268227
View attachment 268228

I did eventually find out where all these GBs of storage is being used.
It seems that undetected chromedriver downloads a chromedriver file in every run.
Code:
/tmp/snap-private-tmp/snap.chromium

View attachment 268229

To remove all temporary chromedrivers
Code:
rm -r /tmp/snap-private-tmp/snap.chromium/*

I did remove all the temp files manually for now.
But I will program this to run automatically.

View attachment 268230


Improvements
I was running with odroid n2l, but changed to beelink eventually.
I was suggested to reboot and clear the cache of chromium browser when I had the memory issues.
But after I have changed this, it took too much more time when rebooting and slowed down the process.

I did order another beelink machine and another lte sim.
So I will be run with 3 machines concurrently soon.
Wish this can speed up the process.
It does come from aliexpress so, it takes like 12 days to arrive here. Xd

Have a nice chill day.
What machine did you bought ? Model and especifications
 
Back
Top