Image Steganography Capacity Improvement Using Cohort Intelligence and Modified Multi-Random Start Local Search Methods

Abstract

In this paper, we have proposed two steganographic techniques which use JPEG compression on greyscale image to hide secret text. JPEG compression is based on discrete cosine transform technique. In order to improve the capacity, a quantization table of size $$16\times 16$$ is practiced instead of a standard JPEG quantization table. Also, the proposed work presents two novel optimization algorithms applied on steganography which are based on the concept of cohort intelligence (CI) with cognitive computing (CC) and Multi-random start local search (MRSLS) algorithm. CI is an emerging optimization algorithm inspired from social learning of one another. This algorithm is being tested to solve unconstrained, constrained and NP-hard combinatorial problems and shows promising results. CC involves self-learning systems and is an emerging area in the field of machine learning. In the proposed work, CI, CC and MRSLS which is inspired from duo-swapping approach and tested to solve NP-hard combinatorial problems, are combined and applied to steganography to produce good results. This work has modified the MRSLS algorithm and applied to steganography to test and validate our results with other comparable algorithms. Experiments are done to test six greyscale images. Experimental results will reveal the quality of stego-images and the secret message embedding capacity.