Is feature selection secure against training data poisoning?

Huang Xiao, Battista Biggio, Gavin Brown +3

Learning in adversarial settings is becoming an important task for application domains where attackers may inject malicious data into the training set to subvert normal operation o…