Towards Evaluating the Robustness of Neural Networks

TL;DR

This paper critically evaluates the robustness of neural networks against adversarial attacks, demonstrating that existing defenses like defensive distillation are ineffective and proposing new, more successful attack algorithms.

Contribution

The paper introduces three new attack algorithms that break defensive distillation and proposes a transferability test to evaluate neural network robustness.

Findings

Defensive distillation does not significantly improve robustness.

New attacks achieve 100% success on distilled and undistilled networks.

Transferability test can effectively break defensive distillation.

Abstract

Neural networks provide state-of-the-art results for most machine learning tasks. Unfortunately, neural networks are vulnerable to adversarial examples: given an input $x$ and any target classification $t$, it is possible to find a new input $x'$ that is similar to $x$ but classified as $t$. This makes it difficult to apply neural networks in security-critical areas. Defensive distillation is a recently proposed approach that can take an arbitrary neural network, and increase its robustness, reducing the success rate of current attacks' ability to find adversarial examples from $95\%$ to $0.5\%$. In this paper, we demonstrate that defensive distillation does not significantly increase the robustness of neural networks by introducing three new attack algorithms that are successful on both distilled and undistilled neural networks with $100\%$ probability. Our attacks are tailored to three distance metrics used previously in the literature, and when compared to previous adversarial example generation algorithms, our attacks are often much more effective (and never worse). Furthermore, we propose using high-confidence adversarial examples in a simple transferability test we show can also be used to break defensive distillation. We hope our attacks will be used as a benchmark in future defense attempts to create neural networks that resist adversarial examples.