Twitter automation test (300 accounts, 6 control groups)

[DropVS]

Well It is basically the same way I automate twitter register.
Automation with undetected chromedriver with semi manual captcha solving.


Appium appdriver is fine, but webdriver is a no go.
It is just like plain selenium, but on phones.
adb is fine. You can automate by finding image match with opencv.

Do you have more details about outlook email creating?
Its better than gmail?

 

[dohyung97022]

Do you have more details about outlook email creating?
Its better than gmail?

Well, there is not much to share,
you don't need to use pyautogui to manipulate clicks tho.

Outlook and gmail is the same,
doesn't matter at all unless you are using some cheap email creator api.

This might help you if you are having problems.
Read this from my first test.

possible red flags on account creation.

1. the email address
I did automate the process with the same emails addresses again and again. Could have a domain trust score.
(@coooooool.com, @popcornfly.com, @coffeetimer24.com etc)
(https://rapidapi.com/calvinloveland335703-0p6BxLYIH8f/api/temp-mail44/).


2. the user-agent
I did not change the user-agent while creating the accounts.
I have doubted this would be a problem.
The user-agent was very common, android phone user agent.
But to be careful, I sould change this if this was the problem.


3. unexpected amounts of traffic.

This is the log of successfully created twitter accounts.
I have created a batch of 300 accounts within the period of 2 days.
They could have detected unusual traffic.


4. Automated process
They could have detected automation when the account was created.
Need to test manualy next time.


5. The chromium browser itself
They might have tricked me, making it look like it has passed.
But It was a fake card all along.

Account creation report

Created the 24 accounts,
the total test cases were not 32 accounts but 24,
cuz I cant do 3x2x2x2 math. XD

Also factor HEADLESS, and MANUAL cannot be done together.
So that is out of the test.

Anyways, these are the factors of the created accounts.

Suspected factor.
1. Create the accounts on a trusted email domain. (bing, google)
2. Change user-agents when created.
3. Create only 32 accounts in a separated timeframe to give less suspicion on traffic.
4. Do it manually.
5. Use the default browser chrome, no selenium attached.

Actual factor.
1. Created by outlook email accounts. (no phone)
2. Created accounts in a 15 minute timeframe.
3. Did manually.
4. Created by new profile in default chrome. (no --remote-debugging-port)
5. 3~4 follows. 2~3 likes. 1 comment. Watch 1 video.
6. 10 minutes of surfing.

Failed factor.
1. Change user-agents
I could not change the user-agent. It was done in default chrome.
I could use a chrome plugin? but did not go in the hassle of doing so.
so all accounts are created with a mac chrome user-agent.

 

[proxygo]

Welcome to test 2 of my twitter automation, a predecessor to TEST 1 (30 accounts).


What is this thread?
So my goal is to conquer twitter marketing.
I have tested before with only 30 accounts as you can see in the link above.
I thought that the number of accounts was too low to prove any point at all.
So I managed to figure out a way to create mass accounts.

Account creation
I have spent the last 6 months of my time trying to automate account creation.
And I have to say that this cannot be automated. (Well, sorta...)

Why?
Arkose funcaptcha is a bitch.

1. Phones does not work. (traffic limit)
Some of you that are knowledged in captcha have advised me to automate with android phones,
Tested that with appium and adb, but still, arkose funcaptcha checks your traffic, and limits you eventually.
My automation tool (proof that I have tried)
View attachment 262256View attachment 262257
about 10 accounts per phone per day (1 account per 2 hour), can be automated.
but I do not have enough phones for mass production.
Maybe when this is profitable I will buy a lot of phones, maybe...

2. Cannot be solved by AI.
These mother fxxxers changes their problems by the day.
I have created an AI for the animal rotation problem, but they keep up with a new question I have never seen before.
I have seen balls with grasses, dices with numbers, trains on chessboards, kids doing jump ropes, icon identification, galaxy identification, people on movie theaters...
New methods just keeps on coming. XD
Some services provides AI for this, but even they cannot keep up with this madness.
All arkose has to do is create a new problem just to fxxx you up.

3. Cannot be solved by other services.
I have tested 2captcha with proxy, but this failed.
2captcha seems to create a new iframe with the secret key, but I think arkose has a way of detecting it.
Did not test all the providers, but I doubt others will work.

Ended up using manpower.
I ended up with semi automation, so I need to keep looking at a screen and solve the captcha manually.
Currently I can create about 40 accounts an hour.
I have 2 LTE proxies with 2 odroid machines running this.
I think I can handle 1 more machine with manpower, so the max would be 60 per hour. (hopefully)
I have other stuff in life, so I will only use an hour per day for creation.

Currently have 229 accounts.
View attachment 262258
I will start testing after the accounts count 300.

I had this problem previously when I take a single action, and the account was banned.
https://www.blackhatworld.com/seo/t...t-4-factors-32-accounts.1461652/post-15942251Well some accounts were locked after register. I had to fix a captcha after that in order to recover it.
(Accounts created after '2023-06-08 00:00:00', lets keep this in mind. This might effect the test.)
All the accounts are not banned, and the actions took in the accounts are valid.
(Tested PM, subscribing to my main account, new tweets are shown.)


So how will I test?
I will have 6 control groups. (50 accounts each)

Conditions
1. automated by undetected chromedriver.
2. useragent is set randomly.
3. Ip is rotated.
4. tested in odroid n2l.
5. All actions are done randomly.

Control groups
1. Like 10, Follow 10, PM 10 per day. (selenium click, types)
2. Like 10, Follow 10, PM 10 per day. (pyautogui click, types)
3. Like 30, Follow 30, PM 30 per day. (selenium click, types)
4. Like 30, Follow 30, PM 30 per day. (pyautogui click, types)
5. Like 50, Follow 50, PM 50 per day. (selenium click, types)
6. Like 50, Follow 50, PM 50 per day. (pyautogui click, types)

The test will go on for 1 month, every day.

What I want to know by this test?
1, Rate limitation of twitter.
2. Is twitter detecting automated key types, or clicks from selenium.
3. Hoping any accounts to survive.

Additional work to do.
1. Code to save all gif logs to s3.
2. Connect datadog to all odroid machines.
3. Create a way to save logs if account is banned, or the action is a success.
4. Create a random action set for automation.

How will I post?
This will be a long journey,
I quit my job, am back in college, so final exam coming up, I won't be posting too soon.
I won't sell my soul on posting threads, last time I had too much fun, and spent too much time on it.
So I wish I keep things simple, no gibberish, only results. (I promise XD)

Hope you have fun.

as you say in the is thread - this will be a long journey
ive asked mods to move this to the Journey section

 

[dohyung97022]

as you say in the is thread - this will be a long journey
ive asked mods to move this to the Journey section

However the mods see fit.

 

[DropVS]

You write:

"
but webdriver is a no go.
It is just like plain selenium, but on phones."

Did you already tried?
Webdriver is more easy to manipulate than appdriver .
Do you know if its possible to change the settings of device with appium? like android id, imei , etc?

 

[dohyung97022]

You write:

"
but webdriver is a no go.
It is just like plain selenium, but on phones."

Did you already tried?
Webdriver is more easy to manipulate than appdriver .
Do you know if its possible to change the settings of device with appium? like android id, imei , etc?

Yes I have tried before.
I have created a register bot in twitter with selenium webdriver.
I have felt the same way. It would be the simplest way. Xd

The results was a instant ban after a single action. (Account creation on test 1)

Appium webdriver does not have any stealth measures. It is like plain selenium but on phones.

Undetected chromedriver or puppeteer extra stealth is selenium that has stealth measures.

 

[DropVS]

Yes I have tried before.
I have created a register bot in twitter with selenium webdriver.
I have felt the same way. It would be the simplest way. Xd

The results was a instant ban after a single action. (Account creation on test 1)

Appium webdriver does not have any stealth measures. It is like plain selenium but on phones.

Undetected chromedriver or puppeteer extra stealth is selenium that has stealth measures.

So in your opinion what its better to Start?
Using undected chrimedriver
Or
Appium with real devices (or emulators..)?

 

[Pidethen]

Interesting, I'm following

 

[dohyung97022]

7/18 ~ 7/22
Test report

Round 1 Test Results (300/300)
Success
Round 2 Test Results (300/300)
13 accounts banned
11 (Set 5) + 2 (Set 4)
Round 3 Test Results (300/300)
3 accounts banned
3 (Set 5)
Round 4 Test Results (300/300)
Success
Round 5 Test Results (300/300)
4 accounts banned
1 (Set 2) + 1 (Set 3) + 1 (Set 4) + 1 (Set 6)
Round 6 Test Results (300/300)
1 (Set 3) + 1 (Set 6)
Round 7 Test Results (300/300)
1 (Set 4) + 3 (Set 5) + 3 (Set 6)
Round 8 Test Results (0/300)
Currently running


Total 29/300 accounts blocked

Thoughts

Bans
Today we have 7 accounts banned.
Set 5 and 6 seems to be getting blocked quickly.

I have heard before that follow / unfollow methods get banned quickly.

I have 100s on of accounts on 3 machines and 0 were banned in 2 years.

For new accounts (2-3months old) - 50 likes/day max which brings 5-8 followers/day
When you gain 3-4k followers you can do 100likes/day which brings 10-15 followers/day
When you exceed 10k followers and 1 year age, you can send likes like crazy. I do 1200 likes/day which brings me 100 followers/day on average

Don't do follow/unfollow it leads to ban very fast. Only likes can be sent without ban risk

I should have added a "Like only" test set.

Currently I would conclude that test set 1 and 2 are the most safest set.
But I should have tried a set with no followers and only likes.
Maybe next time I go under production.

Anyways we can conclude that the more you follow and like, the more likely that your accounts will get banned.
This is of course to be expected. But still, we know where to go from now.
Quantity over quality.
Create as much ids as you can -> Only follow, like a small amount -> rotate to the next id.

(I have no phone varified accounts in this test, this only applies to non pva accounts.)


PM
I have promised that I will do a PM test after a week has passed.
Currently I have added the PM action to my automation.
All the accounts will send 1 pm and will result in a success or a failure.
The results will tell if a week of warmup is sufficient enough for the PM block we have experienced in Round 1.

My third machine has arrived, but I have not set that up yet.
Test results will speed up after the next test.

GLHF in whatever you do.

 

[dohyung97022]

So in your opinion what its better to Start?
Using undected chrimedriver
Or
Appium with real devices (or emulators..)?

undetected chromedriver for automating social media.
Appium, adb with real devices for register.

device automation is very junky in my opinion.
Usb connections and android dev mode keeps going off. Xd
But I have no other options.

Interesting, I'm following

Thanks mate~

 

[dohyung97022]

Oh...

PM is not possible.
Elon blocked me. Xd

It was possible just when I was running the first test.

Currently unverified accounts cannot send PM.

Elon changes things so fast.
People say that threads are the next big thing.
Maybe I will go automate there? XD

 

[DropVS]

What do you mean with that : Usb connections and android dev mode keeps going off. Xd

 

[dohyung97022]

What do you mean with that : Usb connections and android dev mode keeps going off. Xd

The usb connection with your phone and your computer.

Android dev mode (dev options) is what you set up in android in order to connect adb.

 

[DropVS]

The usb connection with your phone and your computer.

Android dev mode (dev options) is what you set up in android in order to connect adb.

Its bad? Or detected ?

 

[dohyung97022]

Its bad? Or detected ?

Not detected.

appium appdriver and adb is just like using the phone itself but by code.
It is impossible for a website to know that you are using it.

 

[Jenson.S]

A lot of phones is possible since there are so many second hand phones out there. This thing is so cool man.

 

[brucesidewalk]

Great experiment. can i know what GUi did you use?

 

[dohyung97022]

Great experiment. can i know what GUi did you use?

Gui?
As you mean user interface?
You mean datadog?
Or my custom program interface?

A lot of phones is possible since there are so many second hand phones out there. This thing is so cool man.

I have 10 used phones to use as register.
But the code is not complete as always. Xd

 

[Jenson.S]

I have 10 used phones to use as register.
But the code is not complete as always. Xd

Creating accounts for me is the most difficult part. I've tried many times, but I can't figure it out in the end.

 

[brucesidewalk]

i got it, you used a watchdog to monitor all actions. so that means, every attempt of account creation, you took a screen shot? or a video? and save it as gif log? and how about proxy server, how did u auto rotate proxies? are you turning of and on the modem or router?