Search papers, labs, and topics across Lattice.
This paper investigates the detection of diffusion-generated time series data, comparing white-box (generator-access) and black-box (signal-only) approaches. They found that a reconstruction-based white-box detector, effective in image domains, fails under generator shift in time series due to the lack of a universal reconstruction prior. Conversely, a simple black-box classifier achieves significantly better performance, with an average F1 of 79.2 and a TPR@1%FPR of 57.2, highlighting the distinct challenges compared to image-based detection.
Diffusion-generated time series are harder to detect than images, as a simple classifier beats a reconstruction-based method that works well for images.
The boundary between real and diffusion-generated time series is becoming increasingly difficult to draw, yet detection in this domain remains underexplored, especially when the generator is unknown. We compare white-box detection, which requires access to the generator, against black-box detection, which operates on the raw signal alone. The white-box approach, a reconstruction-based detector adapted from the image domain, works well in in-distribution but breaks down under generator shift: reconstruction-based detection in images succeeds because large generic generators provide a near-universal reconstruction prior, and no analogous generator exists for time series. In contrast, a simple off-the-shelf classifier used as a black-box detector performs remarkably well, achieving an average F1 of 79.2, a 22.1% relative improvement over the white-box approach, and a TPR@1%FPR of 57.2. Diffusion-generated time series detection is therefore not a direct transfer of the image domain problem. This work provides the first systematic exploration of white-box and black-box detection for diffusion-generated time series. We close by identifying several open and promising directions.
