Enhancement Of Cover Selection – Based Audio Steganography (CAS) Using Block-Based Chaotic Multi – Level Lsb (BCM– LSB) For Balanced Performance

Steganography is a technique in data security by hiding information in an object to avoid
detection. Least Significant Bit (LSB) is one of the audio steganography techniques that
have been proposed to conceal data within an audio file. The weakness of current LSB
technique is low dynamic security. Concurrently, in the audio steganography
fundamental concept, user may select poor cover audio which may reduce capacity,
robustness, imperceptibility, or dynamic security. The objectives of this research are to
design the solution for the audio steganography. This research then proposes a
classification for cover selection methods based on their selection behaviour to identify
strengths and weaknesses of each classification. Cover-quality-based selection is then
selected to be implemented. Block embedding with chaotic map is formulated to
enhance the dynamic security of LSB audio steganography. While to address the cover
selection issues, this research formulated a dynamic security selection criteria and
designed audio-based cover selection method to cater the trade-off between capacity,
robustness, imperceptibility, and dynamic security. A solution known as Coverselection-
based Audio Steganography (CAS) algorithm which consists of Block-based
Chaotic Multi-level LSB (BCM-LSB) and Multi-characteristic-based Cover Audio
Selection (MCAS) is developed based on cover-quality-based selection and evaluated
by incorporating specifically the chaotic block and rounding mechanism for the BCMLSB
and multi-characteristic selection metric and Non-dominated Sorting Genetic
Algorithm 2 (NSGA-II) for the MCAS. Comparing BCM-LSB to other LSB methods
based on seven (7) approaches, its performance in terms of capacity, imperceptibility,
and robustness is comparable individually. However, in terms of dynamic security,
BCM-LSB shows superior performance. For the comparison experiment on trade-off
issues against existing methods that do not consider the trade-off, it is worth noting that
MCAS managed to select set of covers with better score ranges between 66% up to
87%, in which the highest existing method only managed to suggest covers with score
ranges between 47% up to 83%. While for the comparison experiment on the dynamic
security inaccurate metric issue against existing method, MCAS managed to select
covers with higher score ranges between 33% up to 71% while they only managed to
suggest covers with score ranges 34% up to 68%. Lastly, a comparison experiment of
CAS against BCM-LSB using random setting shows that the superiority of CAS as it
managed to select covers with higher score ranges between 28% up to 55% while BCMLSB
using random setting only managed to suggest covers with score ranges 10% up
to 49%. It is concluded that CAS algorithm that consists of BCM-LSB and MCAS
could aid in selecting best cover while considering the trade off in terms of capacity,
imperceptibility, robustness, and dynamic security.