FFT
Fast Fourier Transform — converts an image to the frequency domain where AI models leave detectable signatures.
The Fast Fourier Transform (FFT) decomposes a signal — including a 2D image — into its constituent frequencies. Low frequencies represent smooth gradients and large shapes; high frequencies represent edges, textures, and noise.
Diffusion models leave characteristic patterns in the high-frequency band. Because they generate images by iteratively denoising, the residual structure has a regularity that real cameras don't produce. AI image detectors exploit this by computing the FFT of an image and scoring the high-frequency band for telltale periodicities.
SynthGuard's FFT disruption layer adds carefully calibrated noise to the exact frequency bands detectors score. The added noise has the spectral profile of real sensor noise, so the result looks natural to both detectors and humans. The magnitude is tuned per profile — strict adds more disruption, social adds the minimum needed.
Combined with PRNU injection (which targets spatial-domain signals) the two layers cover both halves of the detector's analysis.
Tools that address FFT
Related terms
Related reading
PRNU, FFT & Sensor Noise — The Forensics Behind Image Authenticity
Image forensics is a small, mathematically dense field that quietly underpins everything from courtroom exhibits to AI detection startups. Three pillars do most of the heavy lifting: PRNU (the sensor…
How AI Image Detectors Actually Work — A 2026 Technical Guide
AI image detectors look magical from the outside — drop an image, get a percentage, ship the verdict. Inside, they are an assembly of brittle statistical signals stacked on top of each other, each ca…