My previous post was deleted because I linked a page, so there might be a small gap for those who didn't read it in time.
@reaaski Funnily enough it was the
@lucky.sparks thread that I stumbled upon when searching for gmail accounts that really triggered that urge to put my black hat on, to anyone who haven't yet his thread is a must read!
@optimalprime hope the following will answer your question.
Really getting carried away with this project and didn't have the time to update thread earlier, but I feel like I have to do it now, not to forget the important details. so here it goes.
Spoiler alert - there's going to be
a lot of scraping.
Handpicking the user(source) accounts seems to be not sustainable, so we have to make a fully autonomous system, heck it's 2022 , with the technology we have access to it definitely is possible, it's only a question of persistence, and that I don't lack.
Will be writing everything in the same order that it happened, so you would also understand that you (at least I) can't get it right the first time always.
We've got a source of about 10k semi-filtered adult accounts, how can we find other related accounts?
Scrape all their posts to find the most frequently used hashtags and sounds, sounds good ay?
Akamai has joined the lobby
Tiktok has many annoying ways of blocking you without actually telling you that you're blocked, say you make a request to /
@Account - everything loads, but it has no posts, okay let's continue?
But you open the very same account with your browser and there it is, a ton of posts
Solution- We work with requests until we get no posts returned from the account, once that happens we fire up our nifty browser, show Akamai that we're the real deal, solve captcha if needed, save the cookies and continue working with requests.
View attachment 219035
Notes, some countries work much better than others.
The captcha get's worse depending on how much you're spamming, starting from the slider one, to the 'select two identical images' and the rotation thing if you really fucked up.
Cool, we've got a fuckton of data scraped
View attachment 219036
Let's analyze what interests us, music and the hashtags.
The top 50~ most frequently used sounds consist of 90% content we are interested in.
And the rest? - Not so much, turns out the sluts also just use whatever sounds are popular, well that's a bummer
Hashtags on the other hand are the real deal, we got at least 500 of hashtags that are primarily used only for adult-ish content.
Notes:
Is an 'accountant' some sort of a slang among younger generation?
About 50 tags are a variation of 'bikinni', I'll leave that one for your judgement.
Sound IDs are incremental, so we could theoretically always check the new sounds, BUT that also goes for any unpopular sound created by a random user.
Plans for the future -
When
finally scraping users videos, get the sound IDs as well, so we can stay up to date with the newest trending sounds/tags in the niche.
Off to scraping those hashtags now!
Pretty much same as before, just with some improvements.
Thankfully when scraping posts from hashtags it shows you some data about the user, like
post count, and user avatar, so we can skip the users with less than 40 videos uploaded AND we can fetch only the users avatar and guess it's gender with some
le'pithon . So we don't need to scrape each users posts individually to decide whether he's/she's hot AND the gender. Saving some requests and traffic, who doesn't like that?
Then the other script takes only females, downloads her video thumbnails and guesses how NSFW the content is.
Notes:
Video thumbnail urls have an expiration token, video thumbnail - 7 days, but user avatar is much shorter. Like 1 day or something don't have it written down, but it's pretty short.
View attachment 219037
DING DONG
A sudden self diagnosis of retardation.
So we are looking aaaalll over the place for NSFW tiktok accounts, and 90% of those accounts are doing OF
And it just so happens that you can actually have a link to your tiktok RIGHT IN YOUR DESCRIPTION
View attachment 219038
Who could have thought that it could be so easy?
Definitely not fucking me.
What does that mean to us?
That's right - more scraping
Not going to get too much in to details of it, but I'm no specialist with OF, and AFAIK it does not allow you to search for users, so I used one of many available search engines.
Ba da beem ba da bow we've got ourselves a hefty list of OF models usernames.
Try to grab the tiktok urls from OF description - quite low percentage of girls have it there, only the advanced bitches with pimps from bhw do. (that's a business opportunity for you right here)
Try to match the usernames on OF to usernames on TT - no dice, very low accuracy and many false positives.
But know what most of them do have?
A link with a wall of all their other links, not sure how it's called, but here's the services ( and for the love of god BHW I'm not promoting these 'services' or selling anything)
allmylinks.com/
username
linktr.ee/
username
my.bio/
username
So what we can do is - bruteforcing all of the above shortners, pretty simple stuff, just take {usernameOnOF} and make a request to all 3 shortners to see if there's a link to tiktok, and grab it.
View attachment 219039
That gets us a couple thousands accounts more.
That still does not satisfy our needs.
Those 50 page 'self improvement' audiobooks my brain has ingested while sleeping start kicking in..
...
Think about why...
...
Think about what...
So I slow down a bit and think why am I doing this.
To get the fresh NSFW
ish content from tiktok
But tiktok doesn't show the newest posts from hashtags like instagram does, or does it?
I do what I do best - a google search.
View attachment 219041
One of those 'how to basic' videos pops up, showing the 'filter by date' working on the mobile app.
My exact reaction to that-
https://www.tiktok.com/@_holubrichard_/video/7106897609121336582
We're reverse engineering tiktok now. I guess
So buckle the fuck up
Artillery- memu,mitmproxy,frida
We're in
View attachment 219042
Spent way longer than I'd wish to admit, fiddling with the request, trying to replicate it but I had to admit that's out of my scope of knowledge.
Googling a bit about the x-gorgon headers only lead me to a few github repositories that both haven't been updated in 2 years~ , both of them claimed being able to generate the gorgon headers for the chinese tt alternative "douyin".
If anyone has any info on that - I'm all ears.
Notes:
If you think whether or not tiktok is tracking your X - the answer is yes. Yes they are.
A couple of hints: timezone, devices uptime, SIM cards MCC,MNC and phone number.
If we do everything properly we can scrape the data without being logged in, just what we wanted!
How it feels to tail -f in windows
View attachment 219044
Notes:
Sometimes on the first start the app likes to crash, just clean the applications data and it's working again!
Of course, stability is an issue with emulators, thank god the recent project I worked with is all about emulators, so it was rather easy to set it up properly and get it working on autopilot.
After some time the IP get's limited, after about 50~ hashtags scraped, just reset the machine, change the proxy IP and you're good to go.
P.s when using mitm as a proxy you can specify --mode upstream:IP
ORT - and the traffic will go out through the supplied proxy.
Cons of this method - I am only able to run 1 emulator at a time, because frida can only connect to one server AFAIK? If you know to specify emulator in frida, or have a tiktoks apk with ssl pinning patched lemme know, it would speed up the operation drastically!
View attachment 219045
So yeah that's like 4 days of work crammed up in to one post, tried to remember all the interesting bits and pieces, hope the memez helped reading it.
If someone want's to play around with those requests, or needs some of the data just ask. IT IS FREE . NOT FOR SALE chill bhw.
Sayonara
