Reversing SynthID

Technical analysis reveals fundamental flaws in Google's SynthID invisible watermarking system for detecting AI-generated images. The encoder leaves detectable artifacts in low-detail areas via histogram analysis, but Gemini's detector inconsistently fails and hallucinates results. These failures undermine SynthID's reliability for media provenance verification.