RealStats: A Rigorous Real-Only Statistical Framework for Fake Image Detection

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zisman, Haim, Shaham, Uri
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908790408871936
author Zisman, Haim
Shaham, Uri
author_facet Zisman, Haim
Shaham, Uri
contents As generative models continue to evolve, detecting AI-generated images remains a critical challenge. While effective detection methods exist, they often lack formal interpretability and may rely on implicit assumptions about fake content, potentially limiting robustness to distributional shifts. In this work, we introduce a rigorous, statistically grounded framework for fake image detection that focuses on producing a probability score interpretable with respect to the real-image population. Our method leverages the strengths of multiple existing detectors by combining training-free statistics. We compute p-values over a range of test statistics and aggregate them using classical statistical ensembling to assess alignment with the unified real-image distribution. This framework is generic, flexible, and training-free, making it well-suited for robust fake image detection across diverse and evolving settings.
format Preprint
id arxiv_https___arxiv_org_abs_2601_18900
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle RealStats: A Rigorous Real-Only Statistical Framework for Fake Image Detection
Zisman, Haim
Shaham, Uri
Computer Vision and Pattern Recognition
Machine Learning
As generative models continue to evolve, detecting AI-generated images remains a critical challenge. While effective detection methods exist, they often lack formal interpretability and may rely on implicit assumptions about fake content, potentially limiting robustness to distributional shifts. In this work, we introduce a rigorous, statistically grounded framework for fake image detection that focuses on producing a probability score interpretable with respect to the real-image population. Our method leverages the strengths of multiple existing detectors by combining training-free statistics. We compute p-values over a range of test statistics and aggregate them using classical statistical ensembling to assess alignment with the unified real-image distribution. This framework is generic, flexible, and training-free, making it well-suited for robust fake image detection across diverse and evolving settings.
title RealStats: A Rigorous Real-Only Statistical Framework for Fake Image Detection
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2601.18900