Be the boss of the BOSS, Break Our Steganographic System!
-


  Home  |   About  |   Rules  |   Materials  |   Credits  |   History  |   Disclaimer  |   'Prizes'  |   Blog  
   Hall of Fame:
Hugobreakers
Binghamton University
Accuracy = 80.3%
Guel&Kurugollu
Queens University
Accuracy = 76.8%
Andreas Westfeld
HTW Dresden
Accuracy = 67%
BossTeam
Czech Tech. Univ.
Accuracy = 65%
View the full list »
BOSSRank

- Testing database of 1000 512x512 greyscale in the pgm format called BOSSRank here (the embedding rate is 0.4bpp and the type of each image (cover or stego) is chosen according to a Bernoulli process with p=0.5).

- Cover database used to generate the Testing database here.

- The ground truth (0 stands for Cover and 1 for Stego) here .

- The mapping for the two cameras used to create BOSSRank (1 is the Leica M9 and 0 the Panasonic Lumix DMC-FZ50 ) here .

- SPAMMER features (aka CDF) here for the testing database.

Reading

- Paper describing the steganographic algorithm called HUGO: Using High-Dimensional Image Models to Perform Highly Undetectable Steganography presented for Information Hiding 2010 , Calgary, Canada.

- Slides and Video of the talk presented by Tomáš Pevný in Information Hiding 2010.

- The description of the BOSS challenge with the rules, the scientific background and the outputs of the contest (to be updated) here .

Training: BOSSBases (v0.93)

- Training database of 9074 512x512 greyscale Cover images (called BOSSBase) in the pgm format here (1).

- Same database with Stego images (embedded at 0.4 bpp) here (1).

- The raw images used to generate the PGM, 7 different cameras, separated in archives of 1000 images here (1).

- The EXIF data of the raw images here .

- SPAM features of Cover images (2nd order, T=3) here and the SPAMMER features here (1).

- SPAM features of Stego images (2nd order, T=3) here and the SPAMMER features here (1).

- The script to generate PGM images from raw images here (1).

- The scripts to generate the stego images from the PGM images for unix here and Windows here .

- The history file of the different versions of BOSSBase here .

Source-codes: HUGO, the embedding algorithm

- Implementations for Unix-based systems, HUGO embedding simulation, SPAM features extraction and tools for using with LibSVM here (2).

- Implementation for Unix based systems, HUGO embedding simulation, compilation using 'configure' here (2).

- Implementation for MacOS, HUGO embedding simulation, compilation using 'configure' here (2).

- Implementations for Windows, HUGO embedding simulation, SPAM features extraction and tools for using with LibSVM here (2).

Source-codes: Feature extractors

- The linux implementation of the SPAM feature is bundled with the implementation of Hugo here (2).

- Implementation for MacOS, SPAM features, compilation using 'configure' here (2).

- Script to extract SPAMMER features (a combination of SPAM and merged features), the current state-of-the-art features, here .

Licences

(1) In order download and use the images provided you must agree to respect the following licence agreement .

(2) Note that by using this implementation, you agree to respect the following software licence.


--