Eigenfaces-Based Steganography

Area of Science:

• Computer Science
• Digital Image Processing
• Information Security

Background:

• Steganography is crucial for covert communication.
• Existing methods often lack robustness or are easily detectable.
• Eigenfaces, derived from Principal Component Analysis (PCA), offer a unique domain for data embedding.

Purpose of the Study:

• To propose a novel transform domain steganography technique utilizing high-order eigenfaces.
• To evaluate the image quality and robustness of the proposed method.
• To establish eigenfaces-based steganography as a viable and secure communication channel.

Main Methods:

• Hiding messages within linear combinations of high-order eigenfaces vectors.
• Utilizing eigenvectors associated with low image variance (high-frequency details).
• Training and evaluation on a large dataset of over 200,000 facial images.

Main Results:

• Image quality remains nearly unaffected after message embedding.
• The eigenfaces-based steganography method demonstrates high robustness against compression attacks.
• The technique is shown to be highly effective for facial images, a common data type on social media.

Conclusions:

• The proposed eigenfaces-based steganography is a promising technique for secure communication.
• It offers superior robustness compared to other state-of-the-art methods, especially against compression.
• The method is reproducible, using publicly available datasets and downloadable implementation.