Rapid advances in generative AI have made it possible for commercial video generation tools to create fabricated surveillance footage. According to a new paper on the arXiv preprint server, such footage can fool both human viewers and automated synthetic video detectors.

The paper introduces CCDF, a benchmark dataset for deepfake detection in surveillance footage. The dataset is intended to support the development and evaluation of detection methods tailored to this specific and high-stakes domain.

Because the abstract notes that these tools are so effective at deceiving existing systems, the authors argue that dedicated benchmarks like CCDF are necessary to keep pace with the threat. The paper does not provide full experimental results in the abstract, but the dataset itself is positioned as a resource for future research.