Learning from Attacks: Attacking Variational Autoencoder for Improving Image Classification

TL;DR

This paper introduces AVIC, a novel framework that uses adversarial attacks on Variational Autoencoders to enhance image classification accuracy beyond traditional methods.

Contribution

It presents a new approach that leverages adversarial attacks on VAEs to improve classification, combining attack and classification networks in a unified framework.

Findings

AVIC achieves higher accuracy than traditional adversarial training.

The framework effectively utilizes implicit information in adversarial examples.

Experimental results demonstrate improved robustness and performance.

Abstract

Adversarial attacks are often considered as threats to the robustness of Deep Neural Networks (DNNs). Various defending techniques have been developed to mitigate the potential negative impact of adversarial attacks against task predictions. This work analyzes adversarial attacks from a different perspective. Namely, adversarial examples contain implicit information that is useful to the predictions i.e., image classification, and treat the adversarial attacks against DNNs for data self-expression as extracted abstract representations that are capable of facilitating specific learning tasks. We propose an algorithmic framework that leverages the advantages of the DNNs for data self-expression and task-specific predictions, to improve image classification. The framework jointly learns a DNN for attacking Variational Autoencoder (VAE) networks and a DNN for classification, coined as Attacking VAE for Improve Classification (AVIC). The experiment results show that AVIC can achieve higher accuracy on standard datasets compared to the training with clean examples and the traditional adversarial training.