How data manipulation could be used to trick fraud detection algorithms on e-commerce sites

A data poisoning attack aims to modify a machine learning model’s training set by inserting incorrectly labeled data with the goal of tricking it into making incorrect predictions.