Membership Inference
A privacy attack where an adversary analyzes a model's output to determine if a specific data record was used in its original training dataset.
Think of It Like This
Like tasting a famous chef's secret soup and confidently knowing if they used your specific brand of tomatoes to make it.
This vulnerability relies on the fact that machine learning models often behave differently (e.g., higher confidence scores) on data they have seen before compared to unseen data. It is a massive privacy risk for models trained on medical or financial records. Differential privacy techniques are commonly used to defend against this attack.