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

arXiv: https://arxiv.org/abs/2605.05224 · HTML: https://mlsanity.com/paper/2605.05224/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 55 references that are in the map.
- First posted on arXiv 18 Apr 2026.

## Builds on (most cited first, 7 of 7)

- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [Bag of Tricks for Training Data Extraction from Language Models](https://mlsanity.com/paper/2302.04460/) (2023)
- [Training Data Extraction From Pre-trained Language Models: A Survey](https://mlsanity.com/paper/2305.16157/) (2023)
- [Learning the Unlearnable: Adversarial Augmentations Suppress Unlearnable Example Attacks](https://mlsanity.com/paper/2303.15127/) (2023)
- [Stable Unlearnable Example: Enhancing the Robustness of Unlearnable Examples via Stable Error-Minimizing Noise](https://mlsanity.com/paper/2311.13091/) (2023)
- [Fooling Adversarial Training with Inducing Noise](https://mlsanity.com/paper/2111.10130/) (2021)
- [A Common Pool of Privacy Problems: Legal and Technical Lessons from a Large-Scale Web-Scraped Machine Learning Dataset](https://mlsanity.com/paper/2506.17185/) (2025)

## Abstract (arXiv)

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.