ConceptAI & ML
Membership inference attack
Determining whether a specific data point was part of a model's training set by measuring the model's loss on that point.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Machine learning models3Language model2Training data2Black-box model access1False positive and false negative rates1Loss score1Privacy auditing1Privacy in machine learning1
Opens up
Concepts that name this one as a prerequisite.
Lecturers
4Each in their own terms, from the easiest level.
Robin Staab
A privacy evaluation approach examining whether specific data samples were part of a model's training set.Moritz Knolle
Attacks on machine learning models where an adversary determines whether a specific data record was part of the training dataset.Ryoto Koike
A task to determine whether a specific text was included in the training data of a target language model.
Lectures that assume it
1Connections
12The lectures that define it or assume it, and the concepts next to it.
Concepts7
- Language model
- Training data
- Adversarial compression ratio
- Aggregate attack success
- Canaries
- Canary forgetting