Deepfake Detection: Emerging Techniques and Evolving Challenges
Krishnaraj Natarajan, A. K. Mathur, Kshitiz Bhargava, Manvendra Singh, Moulik Tejpal · 2024
Several newly developed techniques and tools for manipulating images, audio, and videos have been introduced as an outcome of the recent and rapid breakthroughs in AI, machine learning, and deep learning. While most applications for these techniques or tools are in the fields of entertainment and education, some individuals with unlawful intent have also been benefited from them. These individuals use such techniques for various purposes, including the spread of misleading information and unnecessary propaganda, the incitement of political instability, hate and unrest, as well as for purposes of torture and blackmail. These high-quality and convincing manipulated images, audio, or videos are commonly referred to as ‘Deepfakes’.Since then, various solutions to the problems raised by Deepfakes have been proposed in academic studies. This literature review contains relevant publications that offered a variety of approaches to give an updated summary of the research activities in different types of Deepfake attacks, their detection, and countermeasures. It also assesses the effectiveness of the detection capabilities of different techniques with various datasets and algorithms applied in Deepfake detection, while also outlining the various benefits and drawbacks of various methodologies.