Detection of AI Generated Images Using Combined Uncertainty Measures and Particle Swarm Optimised Rejection Mechanism

TL;DR

A robust detection framework that leverages multiple uncertainty measures to decide whether to trust or reject a model’s predictions, and maintains high acceptance of accurate predictions for natural images and in-domain AI data is presented.

Abstract

As AI-generated images become increasingly photorealistic, distinguishing them from natural images poses a growing challenge. This paper presents a robust detection framework that leverages multiple uncertainty measures to decide whether to trust or reject a model's predictions. We focus on three complementary techniques: Fisher Information, which captures the sensitivity of model parameters to input variations; entropy-based uncertainty from Monte Carlo Dropout, which reflects predictive variability; and predictive variance from a Deep Kernel Learning framework using a Gaussian Process classifier. To integrate these diverse uncertainty signals, Particle Swarm Optimisation is used to learn optimal weightings and determine an adaptive rejection threshold. The model is trained on Stable Diffusion-generated images and evaluated on GLIDE, VQDM, Midjourney, BigGAN, and StyleGAN3, each introducing significant distribution shifts. While standard metrics such as prediction probability and Fisher-based measures perform well in distribution, their effectiveness degrades under shift. In contrast, the Combined Uncertainty measure consistently achieves an incorrect rejection rate of approximately 70 percent on unseen generators, successfully filtering most misclassified AI samples. Although the system occasionally rejects correct predictions from newer generators, this conservative behaviour is acceptable, as rejected samples can support retraining. The framework maintains high acceptance of accurate predictions for natural images and in-domain AI data. Under adversarial attacks using FGSM and PGD, the Combined Uncertainty method rejects around 61 percent of successful attacks, while GP-based uncertainty alone achieves up to 80 percent. Overall, the results demonstrate that multi-source uncertainty fusion provides a resilient and adaptive solution for AI-generated image detection.

Figures (14)

• Figure 1: Overview of the proposed framework. Input images are preprocessed and passed through a feature extractor and task head for classification. Fisher Information, MC Dropout, and GP variance are computed in parallel, normalised, and combined via PSO to produce a unified uncertainty score for accept/reject decisions.
• Figure 2: Pairplots comparing real (blue) and AI-generated (purple) images across four GLCM features: Contrast, Energy, Entropy, and Homogeneity. Each subplot shows the feature distribution and inter-feature relationships for a specific model.
• Figure 3: PSO-derived weighting scheme for combining multiple uncertainty measures from Resnet50.
• Figure 4: Correctly predicted acceptance rates and incorrectly predicted rejection rates across varying classifier probability thresholds, including the threshold $\tau^*$ obtained using Stable Diffusion on ResNet-50.
• Figure 5: Correctly predicted acceptance rates and incorrectly predicted rejection rates across varying classifier total Fisher Information thresholds, including the optimal threshold $\tau^*$ obtained using Stable Diffusion on ResNet-50.