"Outsmarting Deepfake Video"
In March 2022, a deepfake video showed Ukrainian President Volodymyr Zelenskyy ordering his people to lay down their weapons and surrender to Russia. The potential for malicious applications and the rapid evolution of deepfake techniques have ignited a race between the groups producing synthetic media and the scientists seeking more effective and resilient detection methods. Siwei Lyu, a computer scientist at the University at Buffalo, State University of New York, explains that researchers are playing a game of chess in which detection attempts to keep up with or surpass creation. Once adversaries know the techniques used to detect their deepfakes, they can modify their models to render detection algorithms ineffective. In the early days, it was relatively simple for people to identify a fake video due to the prevalence of inconsistencies in skin tone, facial structure, and movement. However, with the advancement of synthesis engines, detection has become increasingly difficult. In order to keep up with the fast development of deepfake technology, researchers have been creating tools to detect telltale indicators of digital forgery. This article continues to discuss the advancement of deepfakes and the efforts that have been made to combat them.