The results obtained by the CREED research project improve the performance in the reversible data hiding research field. The reversible data hiding algorithms allow the insertion of a hidden message into a host digital file. Both the message and the host can be exactly recovered without error by extracting the hidden data. Reversible data hiding in the encrypted domain achieves this objective without compromising the security of the encrypted host file. The research team improve multiple reversible data hiding methods:
data hiding in an encrypted image without preprocessing the image (VRAE vacating room after encryption) using randomly selected pixels was improved, increasing the capacity limit and the encryption security by introducing a block-based permutation stage;
data hiding in an encrypted image using a preprocessing stage(VRBE vacating room before encryption) had its insertion and encryption improved;
data hiding in non-encrypted images using pixel value ordering (PVO), a popular and relevant framework for data hiding was also improved. A flexible vector-based processing order was introduced instead of the existing block-based rigid approach.
New reversible data hiding frameworks were also discovered:
new reversible embedding equations which allow for hidden data insertion by LSB substitution (the hidden data can be directly read without the use of a complicated decoding scheme);
the invisibility of the hidden data was improved by limiting the pixel modifications based on the host image characteristics;
The maximum size of the hidden information that can be reversibly inserted into an image was also increased.
The increase in performance in the encrypted domain was limited, a single proposed methos (VRBE with improved encryption) brings significant improvements over other state-of-the-art schemes. On the other hand, relevant improvements where discovered for classic reversible data hiding on non-encrypted images, a much more competitive research field.
