Pattern Detection in the Activation Space for Identifying Synthesized Content

TL;DR

This paper introduces SubsetGAN, a method that detects GAN-generated images by identifying anomalous activation patterns in neural network layers, outperforming existing detection techniques across various models.

Contribution

The paper presents SubsetGAN, a novel approach that detects synthetic content by analyzing activation subsets in neural networks without prior distribution knowledge.

Findings

Higher detection accuracy than existing methods

Effective across multiple GAN architectures

Works with different proportions of generated content

Abstract

Generative Adversarial Networks (GANs) have recently achieved unprecedented success in photo-realistic image synthesis from low-dimensional random noise. The ability to synthesize high-quality content at a large scale brings potential risks as the generated samples may lead to misinformation that can create severe social, political, health, and business hazards. We propose SubsetGAN to identify generated content by detecting a subset of anomalous node-activations in the inner layers of pre-trained neural networks. These nodes, as a group, maximize a non-parametric measure of divergence away from the expected distribution of activations created from real data. This enable us to identify synthesised images without prior knowledge of their distribution. SubsetGAN efficiently scores subsets of nodes and returns the group of nodes within the pre-trained classifier that contributed to the maximum score. The classifier can be a general fake classifier trained over samples from multiple sources or the discriminator network from different GANs. Our approach shows consistently higher detection power than existing detection methods across several state-of-the-art GANs (PGGAN, StarGAN, and CycleGAN) and over different proportions of generated content.