Minor blind feature based Steganalysis for calibrated JPEG images with cross validation and classification using SVM and
- PDF / 926,524 Bytes
- 20 Pages / 439.37 x 666.142 pts Page_size
- 13 Downloads / 175 Views
Minor blind feature based Steganalysis for calibrated JPEG images with cross validation and classification using SVM and SVM-PSO Deepa D. Shankar 1
& Adresya Suresh Azhakath
2
Received: 11 January 2020 / Revised: 7 August 2020 / Accepted: 2 September 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
The spectacular progress of technology related to the information and communication arena throughout the past epoch made the internet a powerful media for faster communication of data. Though this technology is being admired at one side, there equally exists a challenge for safeguarding the data and privacy of information of a personal without any leak in the data and corresponding mistreatment. Hence, the proposed work primarily aims to investigate the internet communication as well as deter any unwanted happenings, which could occur because of the covert communication. The probable presence of hidden messages is inspected in the digital mass media using the technique of steganalysis. The distinctive features are to be identified, chosen and extracted for universal (blind) steganalysis and are decided by the format of image and its transformation. In this paper, the analysis is carried out in JPEG format images and 10% embedding with 10 fold cross validation. The technique of calibration is used to obtain an estimate of the cover image. Four embedded techniques that have been applied for stegananlysis are Least Significant Bit Matching, LSB Replacement, Pixel Value Differencing (PVD) and F5 respectively. Four different sampling like linear, shuffle, stratified and automatic are considered in this paper. The classifiers used for a comparative study are Support Vector Machine (SVM) and SVM- Particle Swarm Optimization (SVM-PSO). Several kernels namely linear, epanechnikov, multi-quadratic, radial, ANOVA and polynomial are used in classification. The classifier is trained to examine every single coefficient as a separate unit for analysis and the outcome of this analysis helps in finding the decision of steganalysis. Keywords Steganalysis . SVM . SVM-PSO . DCT . PVD . ANOVA . Calibration . Cross validation . LSB matching . LSB replacement
* Deepa D. Shankar [email protected] Extended author information available on the last page of the article
Multimedia Tools and Applications
1 Introduction In recent times, personal data are collected for a number of reasons, including financial transactions, government IDs, military applications and employer licensing and recognition. The protection of many of these data is of paramount importance as the unauthorized use of private information may lead to loss or contamination of information. Even though several strategies are also being designed and implemented, the confidentiality of data is at risk [20]. Cryptography and steganography are the two exclusive and prevalent techniques to secure information. Steganography is considered superior because it makes the message undetectable. In cryptography, the user can know that the message has been
Data Loading...
