REFIT: A Unified Watermark Removal Framework For Deep Learning Systems With Limited Data

@article{Chen2021REFITAU,
  title={REFIT: A Unified Watermark Removal Framework For Deep Learning Systems With Limited Data},
  author={Xinyun Chen and Wenxiao Wang and Chris Bender and Yiming Ding and R. Jia and B. Li and D. Song},
  journal={Proceedings of the 2021 ACM Asia Conference on Computer and Communications Security},
  year={2021}
}
  • Xinyun Chen, Wenxiao Wang, +4 authors D. Song
  • Published 2021
  • Computer Science
  • Proceedings of the 2021 ACM Asia Conference on Computer and Communications Security
Training deep neural networks from scratch could be computationally expensive and requires a lot of training data. Recent work has explored different watermarking techniques to protect the pre-trained deep neural networks from potential copyright infringements. However, these techniques could be vulnerable to watermark removal attacks. In this work, we propose REFIT, a unified watermark removal framework based on fine-tuning, which does not rely on the knowledge of the watermarks, and is… Expand
Removing Backdoor-Based Watermarks in Neural Networks with Limited Data
The Hidden Vulnerability of Watermarking for Deep Neural Networks
RIGA: Covert and Robust White-Box Watermarking of Deep Neural Networks
A Survey on Model Watermarking Neural Networks
SpecMark: A Spectral Watermarking Framework for IP Protection of Speech Recognition Systems
Dataset Inference: Ownership Resolution in Machine Learning
Machine Learning and Hardware security: Challenges and Opportunities -Invited Talk-
...
1
2
3
...

References

SHOWING 1-7 OF 7 REFERENCES
Robust Watermarking of Neural Network with Exponential Weighting
Turning Your Weakness Into a Strength: Watermarking Deep Neural Networks by Backdooring
Adversarial frontier stitching for remote neural network watermarking
Protecting Intellectual Property of Deep Neural Networks with Watermarking
Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks
BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
Overcoming catastrophic forgetting in neural networks