Protect the Intellectual Property of Dataset against Unauthorized Use

TL;DR

This paper introduces a novel method to protect datasets' intellectual property by significantly reducing the accuracy of unauthorized DNN training models, demonstrated on CIFAR-10 and TinyImageNet datasets.

Contribution

The paper presents a new approach for actively protecting datasets from unauthorized use in training deep neural networks, addressing a gap in existing copyright protection schemes.

Findings

Test accuracy drops from 86.21% to 38.23% on CIFAR-10

Test accuracy drops from 74.00% to 16.20% on TinyImageNet

Effective protection against unauthorized dataset use

Abstract

Training high performance Deep Neural Networks (DNNs) models require large-scale and high-quality datasets. The expensive cost of collecting and annotating large-scale datasets make the valuable datasets can be considered as the Intellectual Property (IP) of the dataset owner. To date, almost all the copyright protection schemes for deep learning focus on the copyright protection of models, while the copyright protection of the dataset is rarely studied. In this paper, we propose a novel method to actively protect the dataset from being used to train DNN models without authorization. Experimental results on on CIFAR-10 and TinyImageNet datasets demonstrate the effectiveness of the proposed method. Compared with the model trained on clean dataset, the proposed method can effectively make the test accuracy of the unauthorized model trained on protected dataset drop from 86.21% to 38.23% and from 74.00% to 16.20% on CIFAR-10 and TinyImageNet datasets, respectively.