Summary of The Anatomy Of Adversarial Attacks: Concept-based Xai Dissection, by Georgii Mikriukov et al.
The Anatomy of Adversarial Attacks: Concept-based XAI Dissection
by Georgii Mikriukov, Gesina Schwalbe, Franz Motzkus, Korinna Bade
First submitted to arxiv on: 25 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel study sheds light on the effects of adversarial attacks (AAs) on the concepts learned by convolutional neural networks (CNNs). The researchers employed explainable artificial intelligence (XAI) techniques to analyze the influence of AAs on CNN representations, revealing substantial alterations in concept composition. Specifically, AAs introduce new concepts or modify existing ones, and the perturbation itself can be decomposed into latent vector components responsible for attack success. Notably, these components are target-specific, similar across different AA techniques and starting classes. The study provides valuable insights into the nature of AAs and their impact on learned representations, paving the way for more robust and interpretable deep learning models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers studied how bad attacks can change what neural networks learn. They found that these attacks (called adversarial attacks) make big changes to the ideas or concepts inside the network’s “brain”. These changes are different depending on what they’re trying to trick the network into saying. The study helps us understand how these attacks work and how we can make our networks more reliable. |
Keywords
* Artificial intelligence * Cnn * Deep learning
