Hello guys, ive been working with TikTok traffic for a long time now and lately because of my involvement on DeepVision and Neural networks it reminded me on how TikTok detecting algorithms work and decide to hammer your video into a shadow ban. Would like to know is there any tool which wipes completely the metadata from a video ? So the algorithm cant detect from each platform you took the video earlier and (ex YT or Adult sites) so it will recognize your video as a "new thing", just my assumption. Deep Learning models which do pattern recognition ? Anything to change the pixel order/values so the detector wont be able to recognize the photo by itself ? Bypassing Perceptual Hashing algorithm ? I would be very interested in discussing these.
TikTok - Perceptual Hashing, Deep Learning Models, Metadata... Bypass possible?
- Thread starter xivonidi
- Start date
- Jan 22, 2008
- 5,785
- 8,177
For TikTok it's not 100% possible because their AI.
You break perceptual hash, temporal hash, etc. AI will make voice transcript and make fingerprint.
Solutions for such videos get 50%-70% success. Better to upload to Instagram.
According to photos, here there is not existing solution. Because AI is doing with 100% accuracy with no effort.
You need to change perspective and with another AI change objects on scene.
As much as i've researched they do moderation 30-60 frames per second which yeah its a fast time but as much as i know NNs they still cant operate at 100%, only the human eye can do that, just swap the texture with smth else and see AI go full blind, i may be wrong but thats why i opened this discussion.
PrepareUrself
Newbie
- Aug 27, 2023
- 42
- 4
do you have any informations on how tiktok's perceptual hash works for carousels? i think it should be easier to bypass or am i wrong? what would you need to edit in order to reuse the same batch of photos for different videos? thanks for the thread