SESKStuff Everyone Should Know
SESK: Stuff Everyone Should Know · sesk.info/ai-membership-inference

What does a membership inference attack try to learn?

Whether a specific record was in the training dataset.

Start playing free

  1. The model's programming language
  2. The server's physical location
  3. Whether a record trained the model ✔

Why: The attacker queries a model and studies its outputs. The goal is to infer whether one known record was included during training, which can reveal sensitive participation.

Huh, didn't know that: The original attack worked with black-box access, without seeing internal model parameters.

The dad joke
The record asked the model, do I know you from training?
Like this card?
Play the game

References

Answer: Whether a record trained the model. The attacker queries a model and studies its outputs. The goal is to infer whether one known record was included during training, which can reveal sensitive participation.

  1. Membership Inference Attacks against Machine Learning Models arxiv.org/abs/1610.05820
    determine if the record was in the model's training dataset.
    Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceWhat makes machine learning "supervised"?Artificial IntelligenceWhy can overfitting increase training-data privacy risk?Artificial IntelligenceIs every use of copyrighted work for AI training automatically fair use?

That's stuff everyone should know.