Attacking Visual Language Grounding with Adversarial Examples: A Case Study on Neural Image Captioning

Abstract

Visual language grounding is widely studied in modern neural image captioning systems, which typically adopts an encoder-decoder framework consisting of two principal components: a convolutional neural network (CNN) for image feature extraction and a recurrent neural network (RNN) for language caption generation. To study the robustness of language grounding to adversarial perturbations in machine vision and perception, we propose Show-and-Fool, a novel algorithm for crafting adversarial examples in neural image captioning. The proposed algorithm provides two evaluation approaches, which check whether neural image captioning systems can be mislead to output some randomly chosen captions or keywords. Our extensive experiments show that our algorithm can successfully craft visually-similar adversarial examples with randomly targeted captions or keywords, and the adversarial examples can be made highly transferable to other image captioning systems. Consequently, our approach leads to new robustness implications of neural image captioning and novel insights in visual language grounding.

arXiv commentAccepted by 56th Annual Meeting of the Association for Computational Linguistics (ACL 2018). Hongge Chen and Huan Zhang contribute equally to this work

References in corpus (20)

• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
• Generative Adversarial Text to Image Synthesis
• ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Models
• Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples
• Hierarchical Question-Image Co-Attention for Visual Question Answering
• VQA: Visual Question Answering
• Delving into Transferable Adversarial Examples and Black-box Attacks
• Multimodal Compact Bilinear Pooling for Visual Question Answering and Visual Grounding
• Image Captioning with Semantic Attention
• Deep Visual-Semantic Alignments for Generating Image Descriptions
• Language Models for Image Captioning: The Quirks and What Works
• DeepFool: a simple and accurate method to fool deep neural networks
• Visual Storytelling
• Review Networks for Caption Generation
• Attention Correctness in Neural Image Captioning
• Generating Images from Captions with Attention
• A Convolutional Encoder Model for Neural Machine Translation
• Rich Image Captioning in the Wild
• SentiCap: Generating Image Descriptions with Sentiments
• Video Summarization using Deep Semantic Features

Cited by in corpus (12)

• Large-Scale Adversarial Training for Vision-and-Language Representation Learning
• Towards Fast Computation of Certified Robustness for ReLU Networks
• Characterizing Audio Adversarial Examples Using Temporal Dependency
• Unrestricted Adversarial Examples via Semantic Manipulation
• Defend Deep Neural Networks Against Adversarial Examples via Fixed and Dynamic Quantized Activation Functions
• L2-Nonexpansive Neural Networks
• Robust Sparse Regularization: Simultaneously Optimizing Neural Network Robustness and Compactness
• Query-Efficient Black-Box Attack by Active Learning
• DSRNA: Differentiable Search of Robust Neural Architectures
• Fast Certified Robust Training with Short Warmup
• On Controlled DeEntanglement for Natural Language Processing
• BOSH: An Efficient Meta Algorithm for Decision-based Attacks