Adversarial examples (AEs) pose severe threats to the applications of deep neural networks (DNNs) to safety-critical domains (e.g., autonomous driving and healthcare analytics). While there has been a vast body of research to defend against AEs, to the best of our knowledge, they all suffer from some weaknesses, e.g., defending against only a subset of AEs or causing a relatively high accuracy loss for legitimate inputs. Moreover, most of the existing solutions cannot defend against adaptive attacks, wherein attackers are knowledgeable about the defense mechanisms and craft AEs accordingly.

In this paper, we propose a novel AE detection framework based on the very nature of AEs, i.e., their semantic information is inconsistent with the discriminative features extracted by the target DNN model. To be specific, the proposed solution, namely ContraNet, models such contradiction by first taking both the input and the inference result to a generator to obtain a synthetic output and then comparing it against the original input. For legitimate inputs that are correctly inferred, the synthetic output tries to reconstruct the input. On the contrary, for AEs, instead of reconstructing the input, the synthetic output would be created to conform to the wrong label whenever possible. Consequently, by measuring the distance between the input and the synthetic output with metric learning, we can differentiate AEs from legitimate inputs. We perform comprehensive evaluations under various types of AE attack scenarios, and experimental results show that ContraNet outperforms existing solutions by a large margin, especially for adaptive attacks. Moreover, further analysis shows that successful AEs that can bypass ContraNet tend to have much-weakened adversarial semantics. We have also shown that ContraNet can be easily combined with adversarial training techniques to achieve more outstanding AE defense capabilities.

View More Papers

RamBoAttack: A Robust and Query Efficient Deep Neural Network...

Viet Quoc Vo (The University of Adelaide), Ehsan Abbasnejad (The University of Adelaide), Damith C. Ranasinghe (University of Adelaide)

Read More

RVPLAYER: Robotic Vehicle Forensics by Replay with What-if Reasoning

Hongjun Choi (Purdue University), Zhiyuan Cheng (Purdue University), Xiangyu Zhang (Purdue University)

Read More

DrawnApart: A Deep-Learning Enhanced GPU Fingerprinting Technique

Naif Mehanna (University of Lille, CNRS, Inria), Tomer Laor (Ben-Gurion University of the Negev)

Read More

An In-depth Analysis of Duplicated Linux Kernel Bug Reports

Dongliang Mu (Huazhong University of Science and Technology), Yuhang Wu (Pennsylvania State University), Yueqi Chen (Pennsylvania State University), Zhenpeng Lin...

Read More