ABSTRACT
As machine learning (ML) systems have been dramatically integrated into a broad range of decision-making-sensitive applications for the past years, adversarial attacks and data poisoning attacks have posed a considerable threat against these systems. For this reason, this chapter focuses on two important areas of ML security: adversarial attacks and data poisoning attacks. Specifically, this chapter studies the technical aspects of these two types of security attacks. It has comprehensively described, discussed, and scrutinized adversarial attacks and data poisoning attacks with regard to their applicability requirements and adversarial capabilities. The main goal of this chapter is to help the research community gain the insights and implications of existing adversarial and data poisoning attacks, as well as to increase the awareness of potential adversarial threats when developing learning algorithms and applying ML methods to various applications in the future.
