Conundrum of Deepfakes An Overview and analysis of recent advancements

EasyChair Preprint 5244

Abstract

Recent advancements in artificial intelligence gave rise to deep fakes, which are nowadays more often associated with “Fake news” or “False Information.” Deep fakes involve manipulating the available data to fabricate false data in the form of audio, videos, or photos. The technique can potentially be used with malicious intentions to jeopardize one’s image or establish dominance in global or local politics. There is a significant difference in the amount of expertise available for generating deep fakes and detecting them. More research is going on developing new algorithms and techniques to make deep fakes indistinguishable from real data. While this is good news from the research perspective, at the same time, it is a matter of major concern when it becomes available to those with malicious intentions. This paper mainly emphasizes the societal impact of deep fakes. It presents different analogies to demonstrate how it can interact with people’s mentality to understand why it is important to address it as a problem. The Paper also overviews a few generation techniques as well as the advancements in the field of detection. The Paper also talks about the possible non-algorithmic solutions to tackle the problem.

Keyphrases: Deep Learning for Deepfakes, Deepfakes, Face Re-enactment, Face Swaps, Generation of DeepFakes, Impact of DeepFakes, conundrum of DeepFakes, literature review of DeepFakes

@booklet{EasyChair:5244,
  author    = {Kunal Ghanghav},
  title     = {Conundrum of Deepfakes An Overview and analysis of recent advancements},
  howpublished = {EasyChair Preprint 5244},
  year      = {EasyChair, 2021}}