Adversarial attack application analytics in machine learning

Abstract

Machine learning is one of the most widely studied and applied technologies, but it is itself vulnerable to attack and its algorithms have the risk of privacy leakage. In this article, through the experts currently popular speech recognition scene, reveals how to build the antagonism against data, make its differences with the source data is subtle, so much so that humans can’t through sensory recognition, and machine learning model can accept and the classification of making the wrong decision, at the same time made attack, finally prospects the study model to research the development and application of security and privacy protection.

Keywords

• machine learning
• privacy threats
• adversarial attacks