A Visual Analytics Framework for Adversarial Text Generation

Contributors

Brandon Laughlin, Christopher Collins, Karthik Sankaranarayanan and Khalil El-Khatib

Abstract

This paper presents a framework which enables a user to more easily make corrections to adversarial texts. While attack algorithms have been demonstrated to automatically build adversaries, changes made by the algorithms can often have poor semantics or syntax. Our framework is designed to facilitate human intervention by aiding users in making corrections. The framework extends existing attack algorithms to work within an evolutionary attack process paired with a visual analytics loop. Using an interactive dashboard a user is able to review the generation process in real time and receive suggestions from the system for edits to be made. The adversaries can be used to both diagnose robustness issues within a single classifier or to compare various classifier options. With the weaknesses identified, the framework can also be used as a first step in mitigating adversar- ial threats. The framework can be used as part of further research into defense methods in which the adversarial examples are used to evaluate new countermeasures. We demonstrate the framework with a word swapping attack for the task of sentiment classification.

Acknowledgments

This research was supported by the Communications Security Establishment, the government of Canada’s national cryptologic agency and the Natural Sciences and Engineering Research Council of Canada (NSERC).

Publications

  • B. Laughlin, C. Collins, K. Sankaranarayanan, and K. El-Khatib, “A Visual Analytics Framework for Adversarial Text Generation,” in Proceedings of the IEEE Symposium on Visualization for Cyber Security (VizSec), 2019.
    [Bibtex] [PDF]
    @InProceedings{lau2019a,
      author =    {Brandon Laughlin and Christopher Collins and Karthik Sankaranarayanan and Khalil El-Khatib},
      title =    {A Visual Analytics Framework for Adversarial Text Generation},
      booktitle =   {Proceedings of the IEEE Symposium on Visualization for Cyber Security (VizSec)},
      year = 2019
    }

Research

Guidance in the human–machine analytics process

H-Matrix: Hierarchical Matrix for Visual Analysis of Cross-Linguistic Features in Large Learner Corpora

A Visual Analytics Framework for Adversarial Text Generation

Design by Immersion: A Transdisciplinary Approach to Problem-Driven Visualizations

Semantic Concept Spaces: Guided Topic Model Refinement using Word-Embedding Projections

Discriminability Tests for Visualization Effectiveness and Scalability

Saliency Deficit and Motion Outlier Detection in Animated Scatterplots

ActiveInk: (Th)Inking with Data

Visual Analytics for Topic Model Optimization based on User-Steerable Speculative Execution

ThreadReconstructor: Modeling Reply-Chains to Untangle Conversational Text through Visual Analytics

Detecting Negative Emotion for Mixed Initiative Visual Analytics

EduApps – Supporting Non-Native English Speakers to Overcome Language Transfer Effects

Metatation: Annotation as Implicit Interaction to Bridge Close and Distant Reading

DataTours: A Data Narratives Framework

Perceptual Biases in Font Size as a Data Encoding

Progressive Learning of Topic Modeling Parameters: A Visual Analytics Framework

Abbreviating Text Labels on Demand

NEREx: Named-Entity Relationship Exploration in Multi-Party Conversations

ConToVi: Multi-Party Conversation Exploration using Topic-Space Views

PhysioEx: Visual Analysis of Physiological Event Streams

Using Visual Analytics of Heart Rate Variation to Aid in Diagnostics

Off-Screen Desktop

PivotSlice

Reading Comprehension on Mobile Devices

#FluxFlow: Visual Analysis of Anomalous Information Spreading on Social Media

Balancing Clutter and Information in Large Hierarchical Visualizations

Lexichrome

SentimentState: Exploring Sentiment Analysis on Twitter

Facilitating Discourse Analysis with Interactive Visualization

DimpVis

Glidgets

TandemTable

Simple Multi-Touch Toolkit

Exploring Text Entities with Descriptive Non-photorealistic Rendering

Investigating the Semantic Patterns of Passwords

Bubble Sets: Revealing Set Relations with Isocontours over Existing Visualizations

Parallel Tag Clouds to Explore Faceted Text Corpora

VisLink: Revealing Relationships Amongst Visualizations

DocuBurst: Visualizing Document Content using Language Structure

Tabletop Text Entry Techniques

Lattice Uncertainty Visualization: Understanding Machine Translation and Speech Recognition

WordNet Visualization

// Where the sidebar information is stored
| © Copyright vialab | Dr. Christopher Collins, Canada Research Chair in Linguistic Information Visualization |