Data Poisoning
A malicious attack where adversaries intentionally inject misleading or corrupted samples into a training dataset to degrade the resulting model's performance.
Think of It Like This
Like sneaking fake facts into a student's textbook before an exam so they end up learning the wrong material and failing the test.
This vulnerability occurs when models are trained on user-generated or publicly scraped data that has not been properly vetted. Attackers might introduce subtle perturbations that cause the network to make specific misclassifications at inference time. Defending against data poisoning requires rigorous sanitization and anomaly detection pipelines.