During the Warming phase, it is be provided:
- A training database of 512x512 greyscale Cover images in a raw format (pgm).
- A training database of 512x512 greyscale Stego images (same images with a fixed embedded payload in each).
- The implementation of the embedding algorithm for Unix and Windows.
- The implementation of the SPAM features for Unix and Windows.
- The paper describing the embedding and detection algorithm.
The Ranking phase will start with the disclosure of a test database of 1000 images. Each Stego images will contain a fixed embedded payload of the same size as in the Warming phase.
Rules:
R1 - Ranking phase:
R1.1: One trial will be possible per three day per participant.
R1.2: A trial consists in uploading a file on the BOSS server indicating the predicted class of each image of the test set (stego or cover).
R1.3: Each trial will be evaluated based on a random subset of the images. The score will consequently be only an approximation (it will suffer noise due to the random selection) and will be used to create an informative hall of fame.
R2 - Final ranking:
R2.1: The final ranking will be performed the 15th of December using one trial per participant. The score will be computed using all images of the test set.
R2.2: The winners of the contest will have to prove that their results are only due to the analysis of the images and not to Oracle attacks. If necessary, the organizers will ask them to perform steganalysis on other data-bases for the final ranking.
For any questions, you can contact boss (at) gipsa-lab.grenoble-inp (dot) fr
