J-UNIWARD is an adaptive JPEG-domain steganography method. It assigns costs to changes in JPEG DCT coefficients using a distortion function designed for the JPEG domain.

Quick Summary

Domain:
JPEG.
Type:
adaptive method based on the UNIWARD distortion function.
Tools:
HStego < 0.4, Aletheia simulator, and research implementation in StegoLab.
Main reading:
it is clearly stronger than older JPEG methods, but below J-UNIWARD + Cost Polarization in the current comparison.

Use in tools

J-UNIWARD was used by HStego versions prior to 0.4 for JPEG images. This should be read as a method implemented by that tool version, not as a separate algorithm named “HStego 0.3”.

For detectability experiments, StegoRank also relies on the J-UNIWARD simulator included in Aletheia. A separate J-UNIWARD research implementation is available in StegoLab.

Detectability results

In the evaluated setting, the J-UNIWARD simulator is substantially less detectable than older JPEG tools such as Outguess and F5, and also less detectable than Steghide under the same conditions.

Simulator-based results should be interpreted as method-level results. Embedding simulators are usually slightly harder to detect than real end-user tools such as HStego, because they embed close to the theoretical limit and do not include all practical constraints of a complete tool.

Original paper