A systematic literature review on the effectiveness of deepfake detection techniques
Journal of Cyber Security Technology
Taylor & Francis
School of Science
With technological advances, the generation of deepfake material is now within reach of those operating consumer-grade hardware. As a result, much research has been undertaken on deepfake detection techniques. This work has analysed and measured the performance of various detection techniques using multiple metrics and discussed the effectiveness of these deepfake detection techniques. This has been undertaken by examining and analysing the current state of deepfake detection techniques. Unlike other existing surveys, this work produced a Systematic Literature Review (SLR) on research conducted from the beginning of 2021 to August 2022. This SLR includes tabulated data containing details of the techniques used and the accuracy of those techniques, performance metrics provided in each study, summaries of the datasets used, and challenges and future trends. This SLR has been undertaken with a focus on using a mixed methods approach. This SLR has determined that deep learning (DL) has surpassed machine learning (ML) as the preferred deepfake detection model. However, ML is still a primary focus method in medical imagery. It was also discovered that traditional artificial neural networks are no longer effective and require additional modules to produce ensembled and multi-attentional architectures.