This article deals with adversarial attacks towards deep learning systems for
Natural Language Processing (NLP), in the context of privacy protection. We
study a specific type of attack: an attacker eavesdrops on the hidden
representations of a neural text classifier and tries to recover information
about the input text. Such scenario may arise in situations when the
computation of a neural network is shared across multiple devices, e.g. some
hidden representation is computed by a user's device and sent to a cloud-based
model. We measure the privacy of a hidden representation by the ability of an
attacker to predict accurately specific private information from it and
characterize the tradeoff between the privacy and the utility of neural
representations. Finally, we propose several defense methods based on modified
training objectives and show that they improve the privacy of neural
representations.

本文主要研究基于深度学习自然语言处理中的对抗攻击，探讨攻击者如何通过窃取神经网络文本分类器的隐藏表示来获取敏感信息，为解决此问题，文章提出并论证多个防御方案，改进神经表示的隐私保护性能。

文本隐私保护神经表示

Privacy-preserving Neural Representations of Text

We present a method for transferring neural representations from label-rich
source domains to unlabeled target domains. Recent adversarial methods proposed
for this task learn to align features across domains by fooling a special
domain critic network. However, a drawback of this approach is that the critic
simply labels the generated features as in-domain or not, without considering
the boundaries between classes. This can lead to ambiguous features being
generated near class boundaries, reducing target classification accuracy. We
propose a novel approach, Adversarial Dropout Regularization (ADR), to
encourage the generator to output more discriminative features for the target
domain. Our key idea is to replace the critic with one that detects
non-discriminative features, using dropout on the classifier network. The
generator then learns to avoid these areas of the feature space and thus
creates better features. We apply our ADR approach to the problem of
unsupervised domain adaptation for image classification and semantic
segmentation tasks, and demonstrate significant improvement over the state of
the art. We also show that our approach can be used to train Generative
Adversarial Networks for semi-supervised learning.

提出了一种利用 Adversarial Dropout Regularization 的方法来实现神经表示从有标签的源域到无标签目标域的转移，以提高语义分割和图像分类任务的准确性，并证明了该方法的有效性。

对抗性 Dropout 正则化

Adversarial Dropout Regularization

We investigate the compositional structure of message vectors computed by a
deep network trained on a communication game. By comparing truth-conditional
representations of encoder-produced message vectors to human-produced referring
expressions, we are able to identify aligned (vector, utterance) pairs with the
same meaning. We then search for structured relationships among these aligned
pairs to discover simple vector space transformations corresponding to
negation, conjunction, and disjunction. Our results suggest that neural
representations are capable of spontaneously developing a "syntax" with
functional analogues to qualitative properties of natural language.

通过比较编码器生成的消息向量的真值条件表示与人类生成的指称表达式，我们发现神经表示能够自发地开发出具有功能类似于自然语言定性属性的 “语法”，并从这些对齐的（向量，话语）对中搜索结构化关系，以发现对应于否定、连词和析取的简单向量空间变换。

深度表示中的语言结构类比

Analogs of Linguistic Structure in Deep Representations

Autonomous neural systems must efficiently process information in a wide
range of novel environments, which may have very different statistical
properties. We consider the problem of how to optimally distribute receptors
along a one-dimensional continuum consistent with the following design
principles. First, neural representations of the world should obey a neural
uncertainty principle---making as few assumptions as possible about the
statistical structure of the world. Second, neural representations should
convey, as much as possible, equivalent information about environments with
different statistics. The results of these arguments resemble the structure of
the visual system and provide a natural explanation of the behavioral
Weber-Fechner law, a foundational result in psychology. Because the derivation
is extremely general, this suggests that similar scaling relationships should
be observed not only in sensory continua, but also in neural representations of
``cognitive' one-dimensional quantities such as time or numerosity.

本研究旨在探究如何最优地分布受体，提高自主神经系统在不同统计情况下的处理信息的能力，并在视觉系统的基础上解释行为型魏伯 - 费希纳定律。