EUVIMEDEUVIMED entdecken
Crossref

Machine Learning-Assisted Image Steganography Using Intelligent Pixel Classification and Dynamic LSB Data Embedding

Pankaj Nandan

AI & Cyber Forum: An International Journal · 06.03.2024 · 3 (1) · 04-10

DOI: 10.51470/ai.2024.3.1.04 ↗

Kurz gefasst

Fachartikel in der Zeitschrift AI & Cyber Forum: An International Journal, online erschienen am 08.04.2024. Verfasst von Pankaj Nandan.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The increasing demand for secure digital communication has encouraged the development of intelligent image steganography techniques capable of concealing confidential information while maintaining the visual quality of digital images. Conventional Least Significant Bit (LSB) substitution provides a simple and computationally efficient embedding mechanism; however, uniform pixel selection may introduce unnecessary modifications in smooth image regions and increase the statistical detectability of embedded information. This study proposes a machine learning-assisted image steganography framework that integrates intelligent pixel classification with dynamic LSB data embedding to improve the selection of suitable embedding locations. The proposed approach employs image-derived features, including local variance, gradient magnitude, texture complexity, local entropy, and neighbouring-pixel differences, to classify pixels according to their embedding suitability. A supervised machine learning classifier, such as a Random Forest model, is proposed to estimate the suitability of individual pixels or local image regions for information hiding. Based on the classification output, a dynamic embedding mechanism determines whether a pixel should be excluded from embedding or assigned an appropriate embedding rate. Secret information is subsequently embedded through an adaptive LSB substitution procedure, while a synchronized extraction mechanism enables recovery of the hidden payload. The proposed framework is designed to balance embedding capacity, visual imperceptibility, computational efficiency, and statistical detectability. Its performance is intended to be evaluated using standard image-quality metrics, including Mean Squared Error, Peak Signal-to-Noise Ratio, Structural Similarity Index Measure, embedding rate, and machine-learning-based steganalysis. The study establishes a reproducible methodological framework for investigating the application of supervised learning to adaptive spatial-domain image steganography and provides a basis for future development of intelligent information-hiding systems.

Zitieren

APA

Nandan, P. (2024). Machine Learning-Assisted Image Steganography Using Intelligent Pixel Classification and Dynamic LSB Data Embedding. AI & Cyber Forum: An International Journal, 3(1), 04-10. https://doi.org/10.51470/ai.2024.3.1.04

Vancouver

Nandan P. Machine Learning-Assisted Image Steganography Using Intelligent Pixel Classification and Dynamic LSB Data Embedding. AI & Cyber Forum: An International Journal. 2024;3(1):04-10. doi:10.51470/ai.2024.3.1.04

BibTeX
@article{Nandan2024,
  author = {Nandan, Pankaj},
  title = {Machine Learning-Assisted Image Steganography Using Intelligent Pixel Classification and Dynamic LSB Data Embedding},
  journal = {AI & Cyber Forum: An International Journal},
  year = {2024},
  volume = {3},
  number = {1},
  pages = {04--10},
  doi = {10.51470/ai.2024.3.1.04}
}
Kostenlos starten →