Channel-Level Semantic Perturbations: Unlearnable Examples for Diverse Training Paradigms

TL;DR

This paper investigates the effectiveness of unlearnable examples across different training paradigms, especially pretraining-finetuning, and proposes a hierarchical deception strategy to maintain data privacy.

Contribution

It provides the first systematic analysis of unlearnable examples under pretraining-finetuning and introduces SSC, a novel method to enhance unlearnability in diverse training settings.

Findings

Existing UEs are less effective under pretraining-finetuning.

Frozen shallow layers preserve data semantics, reducing noise impact.

SSC consistently maintains data unlearnability across paradigms.

Abstract

The unauthorized use of personal data in model training has emerged as a growing privacy threat. Unlearnable examples (UEs) address this issue by embedding imperceptible perturbations into benign examples to obstruct feature learning. However, existing studies mainly evaluate UEs under from-scratch training settings, leaving their behavior under the widely adopted pretraining-finetuning (PF) paradigm largely unexplored. In this work, we provide the first systematic investigation of unlearnable examples across diverse training paradigms. Our analysis reveals that loading and freezing pretrained weights significantly weakens the effectiveness of existing UEs methods. We further explain these findings through semantic filtering: while UEs tend to induce models to overfit non-semantic noise, thereby weakening their semantic extraction capabilities, under the PF paradigm, frozen shallow layers preserve data semantics, effectively filtering out distracting information like unlearnable noise. Guided by these insights, we propose a hierarchical deception strategy, Shallow Semantic Camouflage (SSC), that confines the generation process to a semantically valid subspace, aiming to bypass the semantic suppression introduced by pretrained weights. Extensive experiments demonstrate that our method consistently preserves data unlearnability even under challenging training paradigms, such as shallow-layer freezing and semantic-focused pretraining (SF-Pretrain), bridging the critical gap in pretrain-based unlearnable learning.