DeepFake Videos Detection Using Crowd Computing

In recent times, the widespread dissemination of misinformation through realistic artificially generated images and deep faked videos has become a common issue, particularly prevalent on social media platforms targeting popular personalities. Detecting such deep fake content has become increasingly challenging as the technology behind it continues to advance, making it difficult for existing algorithms to avoid false positives. This paper focuses on deepfake detection and makes two key contributions: We propose automated deep fake detection as a supplement to human cognitive abilities, addressing the lack of studies quantifying human capabilities in identifying faked videos from multi-dimensional perspectives. To fill this gap, we conducted experiments with individuals from diverse age groups and professions, comparing their evaluations against automated approaches using state-of-the-art deep fakes. Employing various machine learning algorithms, we predict an individual's ability to correctly classify a video as genuine or fake. By combining human cognitive abilities with video statistical metadata, we achieved an impressive accuracy of 98.3% in predicting participant decisions on video authenticity. To our knowledge, this marks the first attempt to blend these two factors in the context of deepfake detection. © The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management 2023.