Comprehending and Detecting Vulnerabilities using Adversarial Machine Learning Attacks

Abstract

In today’s world, machine learning is an emerging technology which is being used extensively in different domains. In order to offer effective solutions in the broad area of computer security with the use of machine learning (ML) models, applications which identify and protect against potential adversarial attacks are employed. In the ever-growing field of adversarial machine learning, attackers with different extents of accessibility to a machine learning model can launch a number of attacks to achieve their goals. Concurrently, ML models and algorithms are quite susceptible to various cybersecurity threats. In this paper, an in-depth survey has been carried out on the impact of cybersecurity in machine learning and the adversarial attacks which can be encountered in a ML based system.