arXiv:2411.19537v2 Announce Type: replace-cross Abstract: We survey deepfake generation and detection techniques, covering all deepfake media types: image, video, audio and multimodal content. We identify various kinds of deepfakes and construct taxonomies of deepfake generation and detection methods, illustrating the important groups of methods. Next, we gather datasets used for deepfake detection and provide updated rankings of the best performing detectors on the most popular datasets. In addition, we develop a novel multimodal benchmark to evaluate deepfake detectors on out-of-distribution
The rapid advancement and proliferation of generative AI have made deepfake generation and detection a critical and increasingly complex field, necessitating comprehensive surveys to track progress and identify challenges.
This survey provides a critical overview of the state-of-the-art in deepfake technology, which is essential for understanding the evolving landscape of information integrity, cybersecurity, and regulatory frameworks.
The continuous evolution of deepfake techniques necessitates constant updates in detection methods and benchmarks, indicating an ongoing arms race between creators and detectors that impacts trust in digital media.
- · Cybersecurity firms
- · AI ethicists
- · Regulatory bodies
- · Digital forensics
- · Social media platforms (without robust detection)
- · Traditional journalism
- · Public trust in digital media
- · Individuals susceptible to manipulation
The sophistication of deepfake generation will continue to increase, requiring equally advanced detection methods.
Increased deepfake prevalence could lead to greater skepticism towards all digital media, impacting public discourse and verifiable information.
Governments and international bodies may accelerate efforts to legislate and regulate AI-generated content to combat misinformation and manipulation.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG