Inferring Sensitive Attributes from Model Explanations
This work addresses a critical privacy concern for users of machine learning systems by revealing a new vulnerability in model explanations, which is an incremental but important finding in the field of AI security.
The paper tackles the privacy risk of attribute inference attacks on model explanations, showing that adversaries can accurately infer sensitive attributes like race and sex from explanations in various threat models, with attack success rates up to 90% on benchmark datasets.
Model explanations provide transparency into a trained machine learning model's blackbox behavior to a model builder. They indicate the influence of different input attributes to its corresponding model prediction. The dependency of explanations on input raises privacy concerns for sensitive user data. However, current literature has limited discussion on privacy risks of model explanations. We focus on the specific privacy risk of attribute inference attack wherein an adversary infers sensitive attributes of an input (e.g., race and sex) given its model explanations. We design the first attribute inference attack against model explanations in two threat models where model builder either (a) includes the sensitive attributes in training data and input or (b) censors the sensitive attributes by not including them in the training data and input. We evaluate our proposed attack on four benchmark datasets and four state-of-the-art algorithms. We show that an adversary can successfully infer the value of sensitive attributes from explanations in both the threat models accurately. Moreover, the attack is successful even by exploiting only the explanations corresponding to sensitive attributes. These suggest that our attack is effective against explanations and poses a practical threat to data privacy. On combining the model predictions (an attack surface exploited by prior attacks) with explanations, we note that the attack success does not improve. Additionally, the attack success on exploiting model explanations is better compared to exploiting only model predictions. These suggest that model explanations are a strong attack surface to exploit for an adversary.