@user3269078317178 - Somehow posting NSFW to TikTok

[Melow-]

Hi, this account has had roughly 50M views in less than 24 hours posting straight up porn to their page, somehow getting around the filters and landing on peoples FYP's. How is this not being spoken about? Could this be a sort of inside job? There is a link in the bio to an Instagram and a link in the IG bio to an OnlyFans, the traffic they must be getting from this must be insane. Does anyone know how this is possible?

 

[Lukmat]

this is your answer:

@user3269078317178

TikTok is not giving such usernames for around 1 year.
These are very old accounts​

 

[Melow-]

this is your answer:

@user3269078317178​

​

TikTok is not giving such usernames for around 1 year.​

These are very old accounts​

How does that explain completely getting around the adult content filters? Most people cant even post cleavage without it being taken down, these posts show everything and have been up for 24 hours with millions of views. People are also manually reporting them and tiktok are saying they dont violate anything

 

[babayaga14]

Just checked, but how is it possible to have legit pussy pics and tits pics on TT? How is this acc still active?

 

[Melow-]

Just checked, but how is it possible to have legit pussy pics and tits pics on TT? How is this acc still active?

Would love to find out, the traffic the OF must have had in the last 24 hours must've been insane.

 

[babayaga14]

Would love to find out, the traffic the OF must have had in the last 24 hours must've been insane.

Yes, but really curious on how is that possible

 

[Lukmat]

AI SAYS

Fooling image recognition AI while keeping the image indistinguishable to humans is a technique called adversarial attacks. These attacks involve making subtle changes to an image that are usually imperceptible to the human eye but can cause AI models to misinterpret the image entirely. Here’s how this is typically done:

1. Adversarial Perturbations​

• This technique involves adding a small, specially crafted layer of noise to the image. The noise is designed in such a way that a human won't notice the difference, but it can lead the AI model to misclassify the image.
• Gradient-based methods like the Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), and Carlini & Wagner (C&W) attack are popular for generating these perturbations.
• For example, a picture of a cat might have tiny pixel adjustments that make the AI think it's looking at a dog, even though to a human, it still looks like a cat.

2. Adversarial Patches​

• These are visible patterns that, when placed over an image, can cause the AI to misinterpret the entire scene. The patches might look like random designs or symbols but can drastically change the model's output.
• These patches don't need to be large and can be as small as a sticker or a printout. For example, a patch can be stuck on an object or a person, fooling the AI into thinking it’s a completely different object.

3. Universal Adversarial Perturbations​

• Unlike image-specific perturbations, these are single sets of changes that can be applied to multiple images to fool AI across various inputs.
• They work across different contexts and still manage to confuse the AI without being noticeable to humans.

4. Physical Attacks​

• In some cases, adversarial attacks can work in the real world. For example, wearing specially designed glasses or clothing can trick facial recognition systems, making them misidentify or fail to detect the person.
• These techniques have been used to create T-shirts or hats that fool surveillance systems, making people "invisible" to AI.

How It Works​

• Targeting the AI’s Weakness: Most deep learning models process images differently than humans, focusing on small pixel patterns rather than the overall context. Adversarial attacks exploit this sensitivity.
• Optimization Algorithms: Mathematical optimization algorithms are used to find the smallest possible change to an image that still results in misclassification.

Real-World Example​

A famous example is the "stop sign" attack. By adding a few small stickers to a stop sign, researchers made an AI model recognize it as a "yield" sign, even though to the human eye, it still looked like a regular stop sign.

Limitations and Challenges​

• Transferability: An attack that works on one AI model might not work on another unless they are very similar.
• Robustness: AI models can be trained to detect adversarial examples, making them more resistant to such attacks.
• Physical Realism: While digital perturbations are easy to implement, making them work in the physical world is much harder because lighting, angles, and textures can affect the outcome.

Even you see girl, AI can see a tree
Seems TikTok AI is not prepared for that and with few techniques it can be done to fool algorithms

It's 99% this, because he is not using videos.
He is doing photos only

 

[ceoexploit]

Hi, this account has had roughly 50M views in less than 24 hours posting straight up porn to their page, somehow getting around the filters and landing on peoples FYP's. How is this not being spoken about? Could this be a sort of inside job? There is a link in the bio to an Instagram and a link in the IG bio to an OnlyFans, the traffic they must be getting from this must be insane. Does anyone know how this is possible?

i think its an inside job

 

[Lukmat]

i think its an inside job

... and from inside someone is using some stupid CPA? they don't have only fans, just CPA link
You think, someone would make it from inside to earn $10k-$20k? Because probably not more

Also if this would be inside job, he would use videos as well. He is only using images, because he found a way to cloak it from AI

 

[Nessuno]

... and from inside someone is using some stupid CPA? they don't have only fans, just CPA link
You think, someone would make it from inside to earn $10k-$20k? Because probably not more

Also if this would be inside job, he would use videos as well. He is only using images, because he found a way to cloak it from AI

Yeah exactly, that makes no sense being an inside job.

Have seen these accounts popup and he is using image cloaking. Also he is using galleries of 3 images.

 

[Lukmat]

Yeah exactly, that makes no sense being an inside job.

Have seen these accounts popup and he is using image cloaking. Also he is using galleries of 3 images.

I made TensorFlow cloaker and testing now, there are many cloakers based on trained AI models
So he done it by many try&fail attempts
There are so much cloakers and possibilities

 

[Nessuno]

I made TensorFlow cloaker and testing now, there are many cloakers based on trained AI models
So he done it by many try&fail attempts
There are so much cloakers and possibilities

One other thing that he did which is smart is also using the same music, so a lot of creators are like pointing out to that music and then people will be curious and find out, so it makes it even more viral. I had the sound link but seems TikTok deleted it. He started to use more sounds then.

How would he bypass the duplicate content issue since a lot of his posts are the same pics?

 

[Lukmat]

One other thing that he did which is smart is also using the same music, so a lot of creators are like pointing out to that music and then people will be curious and find out, so it makes it even more viral. I had the sound link but seems TikTok deleted it. He started to use more sounds then.

How would he bypass the duplicate content issue since a lot of his posts are the same pics?

My image cloaker failed
Also failed to copy his image and add here

I think there is something else and 100% it's not insider job because TikTok is deleting his accounts.
But these can be also very old accounts and TikTok is just not checking it carefuly.

 

[Kintaroo]

... and from inside someone is using some stupid CPA? they don't have only fans, just CPA link
You think, someone would make it from inside to earn $10k-$20k? Because probably not more

Also if this would be inside job, he would use videos as well. He is only using images, because he found a way to cloak it from AI

is it possible to do this on Snapchat if you find a loophole?

 

[Lukmat]

is it possible to do this on Snapchat if you find a loophole?

Yes

but... I am not even interested to
Well, it's blackhat someway
But I think we shouldn't dig so much and concentrate on that.
I think TikTok will soon ban all his accounts and solve the loophole before we can even find out how it was done.
For experiment I tested and failed.

How much money he could earn? Who knows, but I don't think it's more than $20k-$30k with all his views pointing to single mother account with stupid CPA, it's not only fans. It's pretending to be only fans but it's CPA.
More you will earn focusing on more legit stuff. Even OF can earn a lot on TikTok, I don't know if showing unwanted porn to kids is the way.

I have legit OF clients using my methods and posting on TikTok seducitive videos and makin these $20k-$30k/ month
But they are not spamming porn. People paying for OF are not looking for porn, but for conversation and custom content.
Nowadays even kids can find pages with porn. I don't think so they will go to link and convert his CPA because they saw some pussy.
Actualy, white hat seducitive videos may work even better than naked girls.

 

[ceoexploit]

... and from inside someone is using some stupid CPA? they don't have only fans, just CPA link
You think, someone would make it from inside to earn $10k-$20k? Because probably not more

Also if this would be inside job, he would use videos as well. He is only using images, because he found a way to cloak it from AI

thats a good point

 

[ceoexploit]

My image cloaker failed
Also failed to copy his image and add here

I think there is something else and 100% it's not insider job because TikTok is deleting his accounts.
But these can be also very old accounts and TikTok is just not checking it carefuly.

yeah, TikTok may use weaker AI reviews for older "more verified" accounts.

but why is it that it's only just happening now?

 

[ceoexploit]

Yes

but... I am not even interested to
Well, it's blackhat someway
But I think we shouldn't dig so much and concentrate on that.
I think TikTok will soon ban all his accounts and solve the loophole before we can even find out how it was done.
For experiment I tested and failed.

How much money he could earn? Who knows, but I don't think it's more than $20k-$30k with all his views pointing to single mother account with stupid CPA, it's not only fans. It's pretending to be only fans but it's CPA.
More you will earn focusing on more legit stuff. Even OF can earn a lot on TikTok, I don't know if showing unwanted porn to kids is the way.

I have legit OF clients using my methods and posting on TikTok seducitive videos and makin these $20k-$30k/ month
But they are not spamming porn. People paying for OF are not looking for porn, but for conversation and custom content.
Nowadays even kids can find pages with porn. I don't think so they will go to link and convert his CPA because they saw some pussy.
Actualy, white hat seducitive videos may work even better than naked girls.

this is facts, dont show porn to kids...


i agree that white hate works better because it makes people intrigued and there is a barrier

 

[Nessuno]

yeah, TikTok may use weaker AI reviews for older "more verified" accounts.

but why is it that it's only just happening now?

this is facts, dont show porn to kids...


i agree that white hate works better because it makes people intrigued and there is a barrier

You can use multi quotes bro, just like that ^^

 

[Lukmat]

Yolololo xD
They have probably thousands of such reports.

I don't belive it's not some insider job.
Maybe even moderator from some small country, just marked his accounts as "safe"
and it went viral everywhere