arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2102.07762cs.LGcs.CR
收藏

成员推断攻击在困难问题上更有优势

Membership Inference Attacks are Easier on Difficult Problems

The Hebrew University of Jerusalem(耶路撒冷希伯来大学)

Avital Shafran, Shmuel Peleg, Yedid Hoshen

摘要:

成员推断攻击(MIA)试图检测数据样本是否被用于训练神经网络模型,例如检测版权滥用。我们发现,具有更高维度的输入和输出的模型更容易受到MIA的影响,并在更详细的图像翻译和语义分割,包括医学图像分割模型。我们表明,重建错误可以导致非常有效的MIA攻击,因为它们是记忆的指示。不幸的是,重建误差本身在区分训练中使用的不可预测图像和以前从未见过的易于预测图像方面效果较差。为了克服这一点,我们建议使用一种新的可预测性误差,可以为每个样本计算,其计算不需要训练集。我们的成员误差,通过减去重建误差的可预测性误差,实现高MIA的准确性在大量的基准。

英文摘要:

Membership inference attacks (MIA) try to detect if data samples were used to train a neural network model, e.g. to detect copyright abuses. We show that models with higher dimensional input and output are more vulnerable to MIA, and address in more detail models for image translation and semantic segmentation, including medical image segmentation. We show that reconstruction-errors can lead to very effective MIA attacks as they are indicative of memorization. Unfortunately, reconstruction error alone is less effective at discriminating between non-predictable images used in training and easy to predict images that were never seen before. To overcome this, we propose using a novel predictability error that can be computed for each sample, and its computation does not require a training set. Our membership error, obtained by subtracting the predictability error from the reconstruction error, is shown to achieve high MIA accuracy on an extensive number of benchmarks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑