Evaluating the Effectiveness of Deepfake Video Detection Tools: A Comparative Study
Evaluating the Effectiveness of Deepfake Video Detection Tools: A Comparative Study

Summary/Abstract: This paper focuses on the evaluation of selected tools for the detection of deepfake videos, which pose a growing threat to the integrity of digital information and the trustworthiness of online media. With the increasing availability of artificial intelligence to create highly realistic manipulated content, the need for robust detection systems is important not only in digital forensics, but also in the broader fields of information security and media verification. This study provides a comparative analysis of five deepfake detection tools, including three open source tools (SBI, LSDA, Lipinc) and two commercial solutions (Bio-ID, Deepware), tested on a dataset of 300 manipulated videos from Celeb-DF (v2). The results indicate a better performance of the commercial tools, with Bio-ID achieving a detection accuracy of 98.00% and Deepware 93.47%, outperforming the open source alternatives. The broader implications of this research highlight its potential to strengthen digital trust and combat the spread of disinformation. Reliable detection mechanisms are important for ensuring the authenticity of multimedia content, protecting public figures from attacks on their reputations, and ensuring the credibility of news media. The findings also highlight the importance of continuous innovation in detection algorithms to respond to the evolving sophistication of deepfake technologies. This study provides practical insights for developers, researchers, and policymakers to improve detection tools and contribute to a safer digital environment.

• Details
• Contents