Steganalysis of minor embedded JPEG image in transform and spatial domain system using SVM-PSO

Abstract

Steganalysis recognizes an artifact's expression of a hidden message. The assessment is performed statistically in this paper by extracting the characteristics that show a shift during an embedding. The machine learning technique has been used to classify the stego image and cover image. This paper uses Support Vector Machine with Particle Swarm Optimization (SVM-PSO) as a classifier. A comparative research is carried out using spatial and transform domain steganographic systems. In contrast to the kernel functions, the steganographic systems are Least Significant Bit (LSB) Matching and F5 algorithms. Six various kernel functions and four different sampling methods have been used for evaluation. 10 percent of embedding proportion has been considered for analysis.

Keywords: Steganalysis,F5 algorithms, Sampling methods, Steganographic systems.