Embedding Space Selection for Detecting Memorization and Fingerprinting in Generative Models

TL;DR

This paper investigates how embedding layer choices in Vision Transformers affect memorization detection in generative models, introducing a fingerprinting method that improves identification accuracy of models involved in deepfake generation.

Contribution

It reveals the relationship between layer depth and memorization sensitivity in ViTs and proposes a novel fingerprinting technique based on layer-wise memorization score distributions.

Findings

Memorization scores decrease in deeper layers of ViTs.

Early layers are more sensitive to low-level memorization, later layers to high-level.

The proposed fingerprinting method improves identification accuracy by 30%.

Abstract

In the rapidly evolving landscape of artificial intelligence, generative models such as Generative Adversarial Networks (GANs) and Diffusion Models have become cornerstone technologies, driving innovation in diverse fields from art creation to healthcare. Despite their potential, these models face the significant challenge of data memorization, which poses risks to privacy and the integrity of generated content. Among various metrics of memorization detection, our study delves into the memorization scores calculated from encoder layer embeddings, which involves measuring distances between samples in the embedding spaces. Particularly, we find that the memorization scores calculated from layer embeddings of Vision Transformers (ViTs) show an notable trend - the latter (deeper) the layer, the less the memorization measured. It has been found that the memorization scores from the early layers' embeddings are more sensitive to low-level memorization (e.g. colors and simple patterns for an image), while those from the latter layers are more sensitive to high-level memorization (e.g. semantic meaning of an image). We also observe that, for a specific model architecture, its degree of memorization on different levels of information is unique. It can be viewed as an inherent property of the architecture. Building upon this insight, we introduce a unique fingerprinting methodology. This method capitalizes on the unique distributions of the memorization score across different layers of ViTs, providing a novel approach to identifying models involved in generating deepfakes and malicious content. Our approach demonstrates a marked 30% enhancement in identification accuracy over existing baseline methods, offering a more effective tool for combating digital misinformation.