Efficient ACO Feature selection algorithm for Image stego anomaly detection based on SPAM features
K. Kartheeban, J. Hemalatha, C. Balasubramanian · 2019
The goal of this paper is to enhance the detection accuracy of steganalysis and to reduce the dimensions of high dimensional feature space by removing the irrelevant features. Feature selection is a significant area of research in steganalysis which can greatly inspires the accuracy in detection. In this paper we have proposed an Ant Colony Optimization feature selection (ACOFS- o( n2) for an exploration on the enhancement of revealing (detection) accuracy. ACO is inspired by the artificial ants can pass through on a directed graph. We have chosen the ACO algorithm that can pass through the directed graph with o (2 n) arcs rather than o ( n2) arcs. This greatly reduce the processing time of features and reduced the dimensionality of the feature set extracted from the spatial domain namely Subtractive Pixel Adjacency Matrix (SPAM -686 dimensional), features extracted by steganographic algorithms S-Tools, nsF5, open stego, Hide & Seek. The experimental results show that it significantly improves the detection accuracy and significantly reduces the feature dimensions when comparing with renowned feature selection algorithms.