電気通信システムと管理

Survey on Fake Data Generation and Detection in Telecommunications

Abstract

Moras Haker*

Deep learning advances and the availability of free, large databases have enabled even non-technical people to manipulate or generate realistic facial samples for both benign and malicious purposes. Deep fakes are face multimedia content that has been digitally altered or created synthetically using deep neural networks. The paper begins by describing readily available face editing apps as well as the vulnerability of face recognition systems to various face manipulations. The following section of this survey provides an overview of recent deep fake and face manipulation techniques and works. Four types of deep fake or face manipulations are specifically discussed: identity swap, face re-enactment, attribute manipulation, and entire face synthesis.