MI-LFGOA: multi-island levy-flight based grasshopper optimization for spatial image steganalysis
- PDF / 1,816,521 Bytes
- 28 Pages / 439.37 x 666.142 pts Page_size
- 51 Downloads / 164 Views
MI-LFGOA: multi-island levy-flight based grasshopper optimization for spatial image steganalysis Sonam Chhikara 1 & Rajeev Kumar 1 Received: 4 October 2019 / Revised: 2 July 2020 / Accepted: 9 July 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
A few rich models of steganalysis have been developed that produce high dimensional feature sets for good detection accuracy with high computational cost and time. This paper inspires by the multi-layered concept with a metaheuristic method for reducing the dimensions of feature set and computational cost with maintained detection accuracy. We use Grasshopper Optimization algorithm (GOA) as a baseline for feature reduction, due to its advantage of giving good global optima. While for improving random walk of grasshoppers for balancing local and global solution search in parallel, we use LevyFlight to modify the GOA which is named as Levy-Flight Grasshopper Optimization algorithm (LFGOA). To reduce the general redundant features without affecting detection accuracy, we preprocess the original feature set with PCA followed by LFGOA for further reduction in dimensions of resultant feature set with improved detection accuracy. This same proposed model is executed at different levels (Multi-Islands) with a different set of the population to get better results than single level LFGOA. The proposed framework is named as Multi-Island Levy Flight Grasshopper Optimization (MILFGOA). On the ground of BOSS base 1.01 image database which consists of 10,000 grayscale images, we investigate the proposed steganalysis method for two rich model’s feature sets (34671-D SRM and 686-D SPAM) with different classifiers. The experimental results show that MI-LFGOA is a promising method which achieves 92% to 96% reduction in features dimension while the detection accuracy of the steganalysis is maintained in comparison to other feature selection methods for steganalysis. Keywords Steganalysis . Feature reduction . Feature selection . Optimization
Sonam Chhikara: Deceased on Dec. 16, 2019. This article is dedicated to her memory
* Rajeev Kumar [email protected]
1
School of Computer & Systems Sciences, Jawaharlal Nehru University, New Delhi 110 067, India
Multimedia Tools and Applications
1 Introduction Steganography communicates secretly in digital media by embedding a secret message into media file like image, video, text, and audio files etc. without revealing the existence of such communication inside real digital communication. On the contrary, steganalysis also exists for revealing the steganography which makes embedding more challenging and compel steganography to add an upper level in security. From a long history of ancient steganography to digital steganography, great evolution has been seen with improvement from secret ink to adaptive embedding. In the digital world of steganography, image is preferred over other carrying sources for embedding due to a large number of variations in an image, as small changes are impossible to detect by a
Data Loading...
