Summary of Composite Concept Extraction Through Backdooring, by Banibrata Ghosh et al.
Composite Concept Extraction through Backdooring
by Banibrata Ghosh, Haripriya Harikumar, Khoa D Doan, Svetha Venkatesh, Santu Rana
First submitted to arxiv on: 19 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 The novel Composite Concept Extractor (CoCE) method leverages traditional backdoor attacks to learn composite concepts like “red car” from individual examples, without requiring any labeled training data. By repurposing the trigger-based model backdooring mechanism, CoCE creates a strategic distortion in the manifold of target objects, selectively affecting those with the desired property. Contrastive learning refines this distortion, and a method is formulated for detecting affected objects. Extensive experiments demonstrate the approach’s utility across various datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to learn about complex ideas like “red car” by using individual examples of simpler concepts, like “car” or “red”. The approach uses a trick from computer security to create a special kind of distortion in the data that helps identify objects with specific properties. This method is tested on several datasets and shows promising results. |
