|
2025-01-14 10:14 AEST by Arthur Barrett - Someone has already created an example of how to defeat this:
Breaking Google SynthID Watermarking
Author: Natzir, Technical SEO / Data Scientist
Twitter: @natzir9
Thanks to the work of Zhao, Xuandong, et al., described in their paper "Invisible Image Watermarks Are
Provably Removable Using Generative AI." arXiv preprint arXiv:2306.01953 (2023).
https://gist.github.com/natzir/76e690fd4f5826f3746a109ad08cff16
Here is an alternative AI image detector from HIVE MODERATION:
https://hivemoderation.com/ai-generated-content-detection
added this to chrome on windows 7:
https://chrome.google.com/webstore/detail/hive-ai-detector/cmeikcgfecnhojcbfapbmpbjgllklcbi
But it failed to detect the previous X/Twitter image as AI generated:
25.2% - The input is likely to be: Not-AI-Generated
And again an example of how to defeat it:
Bypass Hive Moderation's AI-Generated Image Detection [Hack]
Author: Natzir, Technical SEO / Data Scientist
Twitter: @natzir9
https://gist.github.com/natzir/8f87bd5e0936c5656f6b9f2858086f2a
|