We present our results in the first BOWS
challenge (Break Our Watermarking System). There were three
given digital photos containing an invisible watermark that
was introduced using informed coding and embedding. The goal
was to remove the watermark from the photos while keeping a
minimum quality of at least 30 dB PSNR (peak signal to noise
We focus on the method used to win the second phase of the
contest with about 58 dB PSNR (18 dB ahead of the best
fellow contributions). This method will be viewed from three
different perspectives: Phase one and two of the contest as
well as with complete knowledge about the implementation and
the secret key.
We introduce a new method to increase the reliability of current steganalytic techniques by optimising the sample order. Space filling curves (e.g., Hilbert curve) take advantage of the correlation of adjacent pixels and thus make the detection of steganographic messages with low change densities more reliable. The findings are applicable, but not limited to LSB steganalysis. An experimental comparison of five different sampling paths reveals that recursive principles achieve by far the best performance. All measures, such as mean distance, median autocorrelation, and the ability to detect even tiny modifications show substantial improvements compared to conventional methods. We elaborate the relationship between those parameters and quantify the effectiveness with a large test database of small images, which are usually hard to
detect. Apart from quantitative advances, visualisation of steganalytic measures can also gain from the application of reverse space filling curves.