Deep Learning-Based Techniques for Identification of Audio DeepFake with Open Issues: A Meta-Analysis
Krity Duhan, Abhishek Kajal · International Journal of Electrical and Electronics Engineering · 2025
Amidst the ongoing advancements in artificial intelligence, a particularly fascinating and alarming progress is the rise of deepfake speech. The emergence of deepfake technology poses a substantial risk to national security, democratic systems, society as a whole and individual privacy. Consequently, it is imperative to create better methods for identifying and mitigating possible deepfake threats. Audio counterfeiting detection is a rapidly developing and important subject. A growing amount of literature is focused on studying deepfake detection computations, which have demonstrated successful results. However, it is important to note that the issue is still far from being completely settled. As synthetic voice generation technology improves, audio deepfake is growing as perhaps the most widespread form of deception. Therefore, the task of differentiating between counterfeit and authentic audio recordings is growing increasingly difficult. Hence, the significance of a system capable of promptly identifying genuine or deceptive audio cannot be exaggerated. In this paper, the evaluation of audio-based deepfake identification methods has been surveyed, and their comparative analysis is being done based on the dataset usage, metrics for evaluation like AUC, EER, a language considered for the dataset taken and the factors such as MFCC, and CQCC. Moreover, the open challenges and future research directions have been highlighted.