Toward Robust Deep Learning Systems against Deepfake for Digital Forensics
Hongmei Chi, Mingming Peng · 2022
In recent years attackers have increasingly adopted Deep Learning (DL) to develop new sophisticated DL-based security attacks or to evade DL-based defense systems. Adversarial attacks are deliberately designed to exploit such vulnerabilities, causing ddep learning models to make a mistake. Deepfakes disrupt crucial digital evidence at military operational decision-making systems. There are only a few forensic methods designed to identifying manipulated media and infer the intention behind them. This Chapter will give a review for design deep forgery images impact in digital forensic and explore various algorithms for helping detect deep forgery images generated by GAN. In addition, we will discuss how to train fairness in DL algorithms to identify the typical features of all the popular GAN algorithms, and how smartphones app can be used to help deepfake detection as well. This Chapter covers the concepts and technology of deepfake forensics. As deep learning technology continues to grow and gain traction many IT professionals are unaware of how deepfakes works but highly interested in its potentiality. The aim of this Chapter is developing an innovated application tool that any digital professional can learn to adopt techniques to detect deepfake development.