Reliable Model Watermarking: Defending against Theft without Compromising on Evasion
Open Access
- 28 October 2024
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 10124-10133
Abstract
With the rise of Machine Learning as a Service (MLaaS) platforms, safeguarding the intellectual property of deep learning models is becoming paramount. Among various protective measures, trigger set watermarking has emerged as a flexible and effective strategy for preventing unauthorized model distribution. However, this paper identifies an inherent flaw in the current paradigm of trigger set watermarking: evasion adversaries can readily exploit the shortcuts created by models memorizing watermark samples that deviate from the main task distribution, significantly impairing their generalization in adversarial settings. To counteract this, we leverage diffusion models to synthesize unrestricted adversarial examples as trigger sets. By learning the model to accurately recognize them, unique watermark behaviors are promoted through knowledge injection rather than error memorization, thus avoiding exploitable shortcuts. Furthermore, we uncover that the resistance of current trigger set watermarking against removal attacks primarily relies on significantly damaging the decision boundaries during embedding, intertwining unremovability with adverse impacts. By optimizing the knowledge transfer properties of protected models, our approach conveys watermark behaviors to extraction surrogates without aggressive decision boundary perturbation. Experimental results on CIFAR-10/100 and Imagenette datasets demonstrate the effectiveness of our method, showing not only improved robustness against evasion adversaries but also superior resistance to watermark removal attacks compared to state-of-the-art solutions.Keywords
All Related Versions
Funding Information
- Fundamental Research Funds for the Central Universities (2242024k30059)
- Natural Science Foundation of Jiangsu Province of China (BK20241272)
- Start-up Research Fund of Southeast University (RF1028623129)
This publication has 41 references indexed in Scilit:
- DepGraph: Towards Any Structural Pruning2023
- Better Trigger Inversion Optimization in Backdoor Scanning2022
- TrojanZoo: Towards Unified, Holistic, and Practical Evaluation of Neural Backdoors2022
- TrivialAugment: Tuning-free Yet State-of-the-Art Data Augmentation2021
- Data-Free Model Extraction2021
- Shortcut learning in deep neural networksNature Machine Intelligence, 2020
- Knockoff Nets: Stealing Functionality of Black-Box Models2019
- Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks2018
- Deep Residual Learning for Image Recognition2016
- ImageNet: A large-scale hierarchical image database2009