JPEG Steganography With Content Similarity Evaluation.

Area of Science:

• Digital Image Processing
• Information Security
• Computer Vision

Background:

• Modern steganalysis utilizes image content similarity, a property often overlooked by existing JPEG steganography techniques.
• Current JPEG steganographic methods primarily focus on content complexity, neglecting the potential of content similarity for improved security.

Purpose of the Study:

• To enhance the undetectability of JPEG steganography by investigating and utilizing image content similarity.
• To develop a novel distortion function that incorporates content similarity for more robust secret data embedding.

Main Methods:

• A new distortion function is designed by updating initial embedding costs of DCT coefficients with DCT block and channel similarities.
• Syndrome Trellis Coding (STC) is employed to embed secret data with minimal distortion using the final distortion function.
• The method processes quantized DCT coefficients and updates embedding costs based on spatial and channel-wise similarities.

Main Results:

• The proposed JPEG steganography scheme demonstrates superior undetectability compared to existing state-of-the-art methods.
• Experimental results validate the effectiveness of incorporating content similarity into the distortion function.
• The method successfully embeds secret data while maintaining a high degree of imperceptibility.

Conclusions:

• Leveraging content similarity, specifically DCT block and channel similarities, significantly improves JPEG steganography undetectability.
• The developed distortion function and embedding strategy offer a more secure approach to hiding data in JPEG images.
• This research advances the field of steganalysis by providing a method that is harder to detect.