Table 10 Comparison of steganographic methods.
From: A deep learning-driven multi-layered steganographic approach for enhanced data security
Approach | Key features | Advantages | Weaknesses |
|---|---|---|---|
Traditional LSB Steganography | Simple embedding in the least significant bits of image pixels | Easy to implement, low computational complexity | Low security, easily detectable by steganalysis tools, limited payload capacity |
Huffman Coding in Steganography | Lossless compression combined with basic steganography techniques | Increases payload capacity and offers basic compression | Limited robustness to attacks, not integrated with deep learning methods for added complexity |
CNN-Based Steganography | Uses CNNs to hide data within images | High robustness against steganalysis, higher payload capacity compared to traditional methods | Computationally expensive, may degrade image quality with large data payloads |
GAN-Based Steganography | GANs for creating steganographic images | Excellent for creating undetectable stegano-images, robust to various steganalysis techniques | GANs are difficult to train, can be prone to instability, and are computationally intensive |
Adversarial Embedding (ADV-EMB) | Embeds data in a way that confuses machine learning-based steganalysis models | Highly resistant to detection by machine learning models | May introduce noticeable artifacts in images, which are complex to implement |
Proposed Multi-Layered Method | Combines Huffman encoding, LSB, and DL for multi-layered security | Efficient lossless compression, high capacity, robust to detection, maintains image quality, uses DL for added security | Requires careful tuning to avoid overfitting, slightly more complex to implement due to the multi-layered approach |
