DeepSearch: A Simple and Effective Blackbox Attack for Deep Neural Networks

2019/10/14 by Fuyuan Zhang, Sankalan Pal Chowdhury, Zhang, Fuyuan +3
Computer Science · #Adversarial Robustness in Machine Learning #Anomaly Detection Techniques and Applications #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.1910.06296

openalex publication_date 2019/10/14 · openalex created_date 2020/08/21 · openalex updated_date 2026/07/28

Abstract

Although deep neural networks have been very successful in image-classification tasks, they are prone to adversarial attacks. To generate adversarial inputs, there has emerged a wide variety of techniques, such as black- and whitebox attacks for neural networks. In this paper, we present DeepSearch, a novel fuzzing-based, query-efficient, blackbox attack for image classifiers. Despite its simplicity, DeepSearch is shown to be more effective in finding adversarial inputs than state-of-the-art blackbox approaches. DeepSearch is additionally able to generate the most subtle adversarial inputs in comparison to these approaches.

Citations

• Gradient-based learning applied to document recognition
• ImageNet Large Scale Visual Recognition Challenge
• Wide Residual Networks
• Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples
• Evolution Strategies as a Scalable Alternative to Reinforcement Learning
• An approach to reachability analysis for feed-forward ReLU neural networks
• A Dual Approach to Scalable Verification of Deep Networks
• On the Limitation of MagNet Defense against L₁-based Adversarial Examples
• Combinatorial Testing for Deep Learning Systems
• Prior Convictions: Black-Box Adversarial Attacks with Bandits and Priors
• On Evaluating Adversarial Robustness
• Decision-Based Adversarial Attacks: Reliable Attacks Against Black-Box Machine Learning Models
• Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples
• Provable defenses against adversarial examples via the convex outer adversarial polytope
• Machine Learning Testing: Survey, Landscapes and Horizons
• Learning Multiple Layers of Features from Tiny Images

Related

• Towards Query Efficient Black-box Attacks: An Input-free Perspective
• Improving Query Efficiency of Black-box Adversarial Attack
• Perceptual Quality-preserving Black-Box Attack against Deep Learning\n Image Classifiers
• Accelerating Targeted Hard-Label Adversarial Attacks in Low-Query Black-Box Settings
• BlackboxBench: A Comprehensive Benchmark of Black-box Adversarial Attacks