Deception by Omission: Using Adversarial Missingness to Poison Causal Structure Learning

Causality-informed machine learning has been proposed as an avenue for achieving many of the goals of modern machine learning, from ensuring generalization under domain shifts to attaining fairness, robustness, and interpretability. A key component of causal machine learning is the inference of causal structures from observational data; in practice, this data may be incompletely observed. Prior work has demonstrated that adversarial perturbations of completely observed training data may be used to force the learning of inaccurate causal structural models (SCMs). However, when the data can be audited for correctness (e.g., it is cryptographically signed by its source), this adversarial mechanism is invalidated. This work introduces a novel attack methodology wherein the adversary deceptively omits a portion of the true training data to bias the learned causal structures in a desired manner (under strong signed sample input validation, this behavior seems to be the only strategy available to the adversary). Under this model, theoretically sound attack mechanisms are derived for the case of arbitrary SCMs, and a sample-efficient learning-based heuristic is given. Experimental validation of these approaches on real and synthetic data sets demonstrates the effectiveness of adversarial missingness attacks at deceiving popular causal structure learning algorithms.