Machine learning plays a significant role in today’s business sectors and governments, in which it is becoming more utilized as tools to help in decision making and automation process. However, these tools are not inherently robust and secure, and could be vulnerable to adversarial modification and cause false classification or risk in the system security. As such, the field of adversarial machine learning has emerged to study vulnerabilities of machine learning models and algorithms, and make them secure against adversarial manipulation. In this paper, we present the recently proposed taxonomy for attacks on machine learning and draw distinctions between other taxonomies. Moreover, this paper brings together the state of the art in theory and practice needed for decision timing attacks on machine learning and defense strategies against them. Considering the increasing research interest in this field, we hope this study provides readers with the essential knowledge to successfully engage in research and practice of machine learning in adversarial environment.
Read more about this research here.