Generative Adversarial Networks (GAN) have received wide attention in the machine learning field for their potential to learn high-dimensional, complex real data distribution. Specifically, they do not rely on any assumptions about the distribution and can generate real-like samples from latent space in a simple manner. This powerful property leads GAN to be applied to various applications such as image synthesis, image attribute editing, image translation, domain adaptation and other academic fields. In this paper, we aim to discuss the details of GAN for those readers who are familiar with, but do not comprehend GAN deeply or who wish to view GAN from various perspectives. In addition, we explain how GAN operates and the fundamental meaning of various objective functions that have been suggested recently. We then focus on how the GAN can be combined with an autoencoder framework. Finally, we enumerate the GAN variants that are applied to various tasks and other fields for those who are interested in exploiting GAN for their research.

Deep neural networks are widely used and exhibit excellent performance in many areas. However, they are vulnerable to adversarial attacks that compromise the network at the inference time by applying elaborately designed perturbation to input data. Although several defense methods have been proposed to address specific attacks, other attack methods can circumvent these defense mechanisms. Therefore, we propose Purifying Variational Autoencoder (Pu-VAE), a method to purify adversarial examples. The proposed method eliminates an adversarial perturbation by projecting an adversarial example on the manifold of each class, and determines the closest projection as a purified sample. We experimentally illustrate the robustness of PuVAE against various attack methods without any prior knowledge. In our experiments, the proposed method exhibits performances competitive with state-of-the-art defense methods, and the inference time is approximately 130 times faster than that of Defense-GAN that is the state-of-the art purifier model.

No abstract

Recent research has shown that a small perturbation to an input may forcibly change the prediction of a machine learning (ML) model. Such variants are commonly referred to as adversarial examples. Early studies have focused mostly on ML models for image processing and expanded to other applications, including those for malware classification. In this paper, we focus on the problem of finding adversarial examples against ML-based PDF malware classifiers. We deem that our problem is more challenging than those against ML models for image processing because of the highly complex data structure of PDF and of an additional constraint that the generated PDF should exhibit malicious behavior. To resolve our problem, we propose a variant of generative adversarial networks (GANs) that generate evasive variant PDF malware (without any crash), which can be classified as benign by various existing classifiers yet maintaining the original malicious behavior. Our model exploits the target classifier as the second discriminator to rapidly generate an evasive variant PDF with our new feature selection process that includes unique features extracted from malicious PDF files. We evaluate our technique against three representative PDF malware classifiers (Hidost'13, Hidost'16, and further examine its effectiveness with AntiVirus engines from VirusTotal.

Generative adversarial networks (GANs) with clustered latent spaces can perform conditional generation in a completely unsupervised manner. However, the salient attributes of unlabeled data in the real-world are mostly imbalanced. Existing unsupervised conditional GANs cannot properly cluster the attributes in their latent spaces because they assume uniform distributions of the attributes. To address this problem, we theoretically derive Stein latent optimization that provides reparameterizable gradient estimations of the latent distribution parameters assuming a Gaussian mixture prior in a continuous latent space. Structurally, we introduce an encoder network and a novel contrastive loss to help generated data from a single mixture component to represent a single attribute. We confirm that the proposed method, named Stein Latent Optimization for GANs (SLOGAN), successfully learns the balanced or imbalanced attributes and performs unsupervised tasks such as unsupervised conditional generation, unconditional generation, and cluster assignment even in the absence of information of the attributes (e.g. the imbalance ratio). Moreover, we demonstrate that the attributes to be learned can be manipulated using a small amount of probe data.