Attacks against ML can be grouped into 3 main families (poisoning, oracle and evasion) and cover the entire lifecycle of an ML-based system.