On Recent Security Issues in Machine Learning

In recent years, applications of machine learning have grown rapidly in various areas. With the accelerating rate of data generation, and recent developments in big data analytics, machine learning has become a de facto standard in many applications that benefited the society in many areas. However, with the increasing number and types of machine learning applications, it has become a target for an increasing number of malicious actors. Security challenges became more complex and diverse in machine-learning-based systems. In this paper, we present a concise survey and discussion of the mechanisms employed by attackers to exploit vulnerabilities in machine learning algorithms or injecting malicious data. The paper focuses on most recent attacks reported in literature and discusses the methods proposed to counter these attacks and reduce their impact.