Think you can spot the difference? The AI can't. Panda or Gibbon? A Beginner's Introduction to Adversarial Attacks is an interactive, beginner-friendly visualization that introduces how machine-learning models can be fooled by malicious adversarial attacks. Built primarily with D3.js and Idyll, my guide focuses on the Fast Gradient Sign Method (FGSM) and shows how tiny, human-imperceptible tweaks to an image can push a ResNet-34 model into making confident mistakes. You can compare clean and subtly perturbed images, explore how these attacks shift model behavior, and examine two versions of ResNet-34, one trained normally and one trained with adversarial methods, to see how they respond differently.
- Accepted and presented at the 7th VISxAI Workshop at IEEE VIS24: VISxAI Workshop Program Info
- Explains adversarial attacks using beginner-friendly interactive visualizations.
- Explores the FGSM attack's impact on ResNet-34 models, with insights into both natural and adversarial images, as well as standard and adversarial trainings.
- Includes embedding-level and instance-level analysis to show how adversarial perturbations affect models.
Presentation at the VISxAI Workshop
Video of my presentation at the 7th VISxAI workshop (starts from 1:30:44)
Explore the live VISxAI demo.
Watch the walkthrough on YouTube.
- Python
- PyTorch
- t-SNE
- Machine Learning
- Computer Vision
- Adversarial Machine Learning
- XAI Visualization
- D3.js
- Idyll-lang
- Adversarial Machine Learning
- FGSM Attack
- Adversarial Attack
- Image Classification
- Visualization
- ResNet
- Model Robustness
Yuzhe You, Jian Zhao