From: Digital image watermarking: its formal model, fundamental properties and possible attacks
Models in use | Objectives | Inputs and outputs | Component functions | Underlying theory | Limitations |
---|---|---|---|---|---|
Jian and Koch et al. [5] | To describe digital watermarking schemes | Original data | Embedding (bit-carrier selector, bit-noise generator, bit-carrier modifier) | Steganography | Limited consideration of the inputs, outputs, component-functions, and watermarking properties for image applications |
 |  | Watermarked data | Extraction (bit-carrier selector, bit-pattern matching) | Spread-spectrum communications | Limited to spread-spectrum communication-based watermarking schemes |
 |  | Degraded data (as a copy of watermarked data) |  | Signal processing |  |
 |  | Identification code (as watermark) |  |  |  |
O’Sullivan et al. [8] | To determine the optimal hiding strategy, where watermarking is considered as a game between an attacker and information hider | Input and output data (e.g., images, audio, etc as a vector) | Encoder | Information theory | Limited consideration of the image application scenarios, inputs, outputs, component-functions, and watermarking properties |
 |  | Message (as watermark) | Decoder | Steganography |  |
Cox et al. [11] | To examine the similarities between watermarking and traditional communication models | Cover data (as a vector) | Perceptual distance function | Spread-spectrum communications | Limited consideration of image application scenarios (e.g., that use only spread-spectrum-based schemes), inputs, outputs, component functions, and watermarking properties |
 |  | Watermark message | Encoding function |  |  |
 |  | Watermarked cover data | Extraction function |  |  |
 |  |  | Mixing function |  |  |
Petitcolas et al. [4] | To illustrate a simplified case of watermarking concept | Mark (as fingerpring or watermark) | Embedding | Information hiding | Limited consideration of inputs, outputs, and components |
 |  | Stego image | Recovery |  | May not be useful to study image watermarking schemes rigorously |
 |  | Marked image |  |  |  |
Cohen and Lapidoth [9] | To compute the coding capacity of the watermarking game for a Gaussian cover text and squared mean error distortions | Cover text | Encoder | Game theory | Limited consideration of inputs, outputs, and components |
 |  | Message | Decoder | Information theory | Watermarking is considered as a game in a copyright application scenario |
 |  | Stego-text |  |  |  |
 |  | Secret key |  |  |  |
Adelsbach et al. [12] | To analyse security of watermarking schemes against protocol attacks (e.g., copy, ambiguity attacks) | Unwatermarked object | Key generation | Cryptography | Limited consideration of inputs, outputs, and components |
 |  | Watermarked object | Embedding |  | Application scenarios are limited to dispute resolving protocols |
 |  | Watermark | Detection |  |  |
 |  | Key |  |  |  |
Barni et al. [13] | To provide a general security framework for robust watermark | Original content | Embedding (feature extraction and mixing, watermark generation) | Information theory | The concept of fair and unfair attacks may not be realistic |
 |  | Watermark | Decoding | Cryptography | Limited consideration of inputs, outputs, and components (e.g., what original content includes) |
 |  | Watermarked content |  | Signal processing |  |
 |  | Key (for embedding and detection) |  |  |  |
Li et al. [1] | To illustrate the formulation of the security definitions and the attacker models | Original and watermarked work (as a vector) | Watermark generation | Data hiding | Limited consideration of inputs and outputs |
 |  | Watermark (as bit sequence) | Watermark embedding | Cryptography | The model represents only a simplified case of watermarking |
 |  |  | Watermark detector | Signal processing |  |
 |  |  | Perceptual distance function |  |  |
Moulin and O’Sullivan [15] | To evaluate hiding capacity in an optimal attack context (as a data-hiding game) | Host-data (image, audio, video, etc.) | Encoder | Information theory | Limited consideration of inputs, outputs, and component functions (e.g., inputs and outputs are not conventional for watermarking) |
 |  | Message | Decoder | Data-hiding |  |
 |  | Side information |  | Game theory |  |
 |  | Composite data (contains hidden message) |  |  |  |
Mittelholzer [6] | To characterize embedding process and attacked stego image (for analyzing secrecy and robustness in terms of mutual information) | Cover data | Stego encoder | Information theory | Limited consideration of inputs, outputs, and component functions |
 |  | Key | Stego channel | Steganography | More related to steganography schemes |
 |  | Secret message | Stego decoder |  |  |
Cachin [10] | To quantify steganographic security | Cover text | Key generation | Information theory | Limited consideration of inputs, outputs, and component functions |
 |  | Stego text | Embedding | Steganography | More related to steganography schemes |
 |  | Secret key | Extraction |  | Limited to the passive attack scenarios |
Adelsbach et al. [7] | To formalize robustness considered as a core security property, of watermarking | Cover data | Key generation | Cryptography | Limited consideration of inputs, outputs, and component functions |
 |  | Stego data | Embedding | Limited to robust watermarking schemes |  |
 |  | Watermark | Detection |  |  |
 |  | Key (for embedding and detection) |  |  |  |
 |  | Secret parameter (used as key generation input) |  |  |  |