Robust Learning for Data Poisoning Attacks
Robust Learning for Data Poisoning Attacks
复制标题
DOI：
--
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Yunjuan Wang;Poorya Mianjy;R. Arora
中科院分区：
文献类型：
--
作者：
Yunjuan Wang;Poorya Mianjy;R. Arora
我们研究了随机逼近方法对数据中毒攻击的鲁棒性。我们专注于具有ReLU激活的双层神经网络，并表明在无限宽度网络诱导的RKHS中的可分性的特定概念下，具有随机梯度下降的训练（有限宽度）网络对数据中毒攻击具有鲁棒性。有趣的是，我们发现，除了一个下限的网络的宽度，这是在文献中的标准，我们还需要一个distributiondependent上界的宽度强大的推广。我们提供了广泛的实证评估，支持和验证我们的理论结果。
We investigate the robustness of stochastic approximation approaches against data poisoning attacks. We focus on two-layer neural networks with ReLU activations and show that under a specific notion of separability in the RKHS induced by the infinite-width network, training (finitewidth) networks with stochastic gradient descent is robust against data poisoning attacks. Interestingly, we find that in addition to a lower bound on the width of the network, which is standard in the literature, we also require a distributiondependent upper bound on the width for robust generalization. We provide extensive empirical evaluations that support and validate our theoretical results.