Summary of Holmes: to Detect Adversarial Examples with Multiple Detectors, by Jing Wen
HOLMES: to Detect Adversarial Examples with Multiple Detectors
by Jing Wen
First submitted to arxiv on: 30 May 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
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 Medium Difficulty summary: A novel approach to defend deep neural networks (DNNs) from adversarial attacks is proposed, focusing on identifying potential threats at the output level rather than modifying the DNN models themselves. The authors introduce HOLMES (Hierarchically Organized Light-weight Multiple dEtector System), a system that combines multiple detectors to detect unseen adversarial examples with high accuracy and low false positive rates. This approach is more effective than single detector systems and can be used in conjunction with other defenses. The paper presents two methods for training dedicated detectors for each label, as well as top-k logits, ensuring diversity and randomness in the detection process. HOLMES is compatible with various learning models and external APIs, making it a valuable tool for practitioners. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty summary: Imagine trying to trick a computer by adding tiny changes to images, but the computer still gets them wrong. This is called an “adversarial attack” and can be very harmful. The researchers propose a new way to detect these attacks before they cause damage. They introduce a system called HOLMES that uses multiple detectors to identify potential threats and prevent them from causing harm. This approach is more effective than previous methods and can be used in combination with other defenses to keep computers safe. |
Keywords
» Artificial intelligence » Logits
