Efficient Visualization of Neural Networks with Generative Models and Adversarial Perturbations

TL;DR

This paper introduces a simplified deep visualization method using a generative model with a discriminator acting as a guide, capable of producing class-specific images and generating effective adversarial examples with high fooling rates.

Contribution

It presents a novel, less complex model for neural network visualization that also serves as an adversarial attack technique, linking interpretability and vulnerability assessment.

Findings

Achieves up to 94.5% fooling rate in adversarial attacks.

Outperforms traditional adversarial methods in targeted and non-targeted attacks.

Uses a non-adversarial training process with a simplified architecture.

Abstract

This paper presents a novel approach for deep visualization via a generative network, offering an improvement over existing methods. Our model simplifies the architecture by reducing the number of networks used, requiring only a generator and a discriminator, as opposed to the multiple networks traditionally involved. Additionally, our model requires less prior training knowledge and uses a non-adversarial training process, where the discriminator acts as a guide rather than a competitor to the generator. The core contribution of this work is its ability to generate detailed visualization images that align with specific class labels. Our model incorporates a unique skip-connection-inspired block design, which enhances label-directed image generation by propagating class information across multiple layers. Furthermore, we explore how these generated visualizations can be utilized as adversarial examples, effectively fooling classification networks with minimal perceptible modifications to the original images. Experimental results demonstrate that our method outperforms traditional adversarial example generation techniques in both targeted and non-targeted attacks, achieving up to a 94.5% fooling rate with minimal perturbation. This work bridges the gap between visualization methods and adversarial examples, proposing that fooling rate could serve as a quantitative measure for evaluating visualization quality. The insights from this study provide a new perspective on the interpretability of neural networks and their vulnerabilities to adversarial attacks.