A dversarial machine learning (AML) is a recent research field that investigates potential security issues related to the use of machine learning (ML) algorithms in modern artificial intelligence (AI)-based systems, along with defensive techniques to protect ML algorithms against such threats. The main threats against ML encompass a set of techniques that aim to mislead ML models through adversarial input perturbations. Unlike ML-enabled crimes, in which ML is used for malicious and offensive

more »

... rposes, and ML-enabled security mechanisms, in which ML is used for securing existing systems, AML techniques exploit and specifically address the security vulnerabilities of ML algorithms. Consider, for example, an automatic surveillance camera that uses certain ML algorithms. The system monitors people entering and leaving a building in real time. A person wearing a special T-shirt walks by the building, but the camera does not detect the person's presence, as the T-shirt has a special pattern that effectively conceals the person from the camera. Such a pattern can be constructed and optimized against the target system by leveraging attack algorithms developed in the AML research field. 1