Article: Understanding ML Model Poisoning: How It Happens and How to Detect It

Researchers have detailed how data poisoning techniques like label flipping and backdoor injections can compromise machine learning systems. These methods manipulate training datasets to force models into making incorrect predictions or exhibiting hidden malicious behaviors.