Watermarking 3D Models

Guided by:
Jyothi Vijayan Presented by:
Assistant Professor
Computer Science Dept. Vaisakh V
College of Engineering Adoor ADR20CS051
Contents
1. Introduction
2. Digital watermarking
3. Characteristics of digital watermarking
4. Applications
5. Attacks against watermarking
6. Watermarking techniques
7. Digital image
8. 3D image
9. Implementation technologies
10. Double layer watermarking scheme
11. Implementation of the method
12. Performance evaluation
13. Future scope
14. Conclusion
Introduction
● In today’s world, both information technology and three–dimensional (3D)
models have seen a massive surge in popularity.
● With this popularity comes a need for protection, whether it is the protection of
sensitive data from being stolen or the protection of 3D models against
copyright violations and theft, and that could be achieved through digital
watermarking.
● The thesis proposes “Double Layer Watermarking scheme for 3D Models
Using Full LSB and Chaotic Functions” as a comprehensive solution to
protect 3D model integrity and embedded sensitive information.
What is digital watermarking?
Digital watermarking refers to the process of embedding information into digital
media (like images, audio, video, or 3D models) to signify ownership, authenticity,
or to convey additional data without altering the media's perceptual quality.

Purpose:

It serves as an invisible identifier, allowing content owners to claim ownership,

track copyright, verify authenticity, or embed additional information within the
media.
Digital watermarking - continuation
Digital watermarking is mostly made up of two main modules, those being the
watermark embedding module and the watermark detection and extraction module.
Characteristics of digital watermarking
● Transparency: Watermarks must be imperceptible to the human eye, preserving the
original media's quality.
● Robustness: They should withstand various attacks and modifications, requiring the
original embedding technique for alteration.
● Capacity: The embedded bits must be sufficient for later detection without
compromising the media's integrity.
● Security: Only authorized individuals with specific knowledge or keys should access or
modify the watermark.
● Complexity: Balancing high security with practical usability in the time taken for
embedding and extraction processes.
Applications of digital watermarking
● Copyright Protection: Embedding watermarks to identify and credit content owners, aiding in
settling copyright disputes.
● Copy Control: Using watermarks to inhibit unauthorized copying by including copyright regulations
within the content.
● Fingerprinting: Tracing illegal copies back to their source by embedding unique watermarks in
distributed content, aiding in identifying responsible users.
● Authentication: Using fragile or semi-fragile watermarks to verify content integrity by comparing
original and target content watermarks for alterations.
● Secret Communication: Embedding secret messages within digital content for later decryption by
intended recipients.
● Medical Applications: Protecting patient information through watermarking to ensure privacy and
avoiding critical errors by visibly watermarking patient names on medical reports and scans.
Attacks against watermarks
● Signal Attacks: Various forms such as Gaussian noise, histogram equalization, or blur
attacks aim to distort the watermark or content without focusing on the watermark's encryption
method.
● Cryptographic Attacks: Concentrate on breaking encryption techniques or keys used during
watermark embedding, including brute force and oracle attacks seeking to remove or modify
the watermark.
● Geometric Attacks: Alter digital content through geometric operations like scaling, rotation, or
cropping, often damaging the watermarked image itself, as seen in mosaic attacks.
● Protocol Attacks: Split into invertible attacks, where the attacker discovers their own
watermark, compromising the owner's identification, and copy attacks, where the attacker
replicates watermarks across different digital media to assert ownership.
Watermarking techniques

● Spatial Domain Techniques: Utilize methods like LSB, pixel value

differencing, predictive coding, etc., to manipulate pixel values for embedding
watermarks.
● Transform Domain Techniques: Employ DCT, DFT, DWT, SVD to transform
and embed watermarks within frequency components, offering better
imperceptibility, security, and robustness.
Digital Image
● A digital image functions as a mathematical equation involving x and y
variables, with 'a' denoting brightness at each (x, y) point.
● Images are often viewed in specific sections or regions of interest, rather than
as a whole picture, allowing for focused processing.
● A digital image comprises a 2D grid of pixels, holding color data; 8 bits for
grayscale and 24 bits for RGB (Red, Green, Blue) colored images.
3D Image
A 3D image, similar to a 2D image, uses voxels (like pixels but in 3D space).
3D image - continuation
When it comes to constructing 3D images, one of the methods used is slicing:

● Slicing Technique: This method builds a 3D model by merging multiple 2D slices of the object,
each offering a different view or perspective. After merging, adding light enhances the object's
appearance, providing depth. This technique divides the 3D image into 2D slices for processing.
Implementation of watermarking for 3D models (Techniques)

Wolfram Mathematica:
● Advantages:
● Vast array of built-in mathematical functions and libraries, suitable for
processing 3D models.
● Excellent visualization capabilities for both 2D and 3D graphics.
● High precision in computational results.
● Disadvantages:
● Slow interpreter, resulting in longer compilation times.
● Steeper learning curve for beginners.
● Limited sharing capabilities for central packages and design tools.
Implementation of watermarking for 3D models (Technologies)

Mathworks MATLAB:
● Advantages:
● Ease of use with a command-line interface and simple language.
● Platform independence and a comprehensive toolbox for various
functions.
● Good for generating quick prototypes and simulations.
● Disadvantages:
● Slower compilation due to interpreted language.
● High cost of acquisition and system requirements.
Implementation of watermarking for 3D models (Technologies)

Python:
● Advantages:
● Simple, readable syntax resembling plain English.
● Open-source nature and extensive library availability.
● Platform independence and ease of use.
● Disadvantages:
● Slower compilation compared to compiled languages.
● Higher memory consumption and runtime errors during testing.
Implementation of watermarking for 3D models (Technologies)

Optimal Choice:

Wolfram Mathematica is the chosen technology for implementing the

watermarking scheme due to its exceptional 3D processing capabilities, vast
mathematical functions and precision.
Double Layer Watermarking scheme for 3D Models

Slice-Based Approach:
● 3D model sliced into 2D segments for individual processing.
● Enables independent handling of each slice for data embedding.
Encryption Layer 1:
● Message undergoes encryption using a shared chaotic function-generated key.
● Ensures secure transformation of the message for transmission.
Encryption Layer 2 - Embedding Process:
● Utilizes Full LSB (Least Significant Bit) embedding technique.
● Embeds the encrypted message within 2D slices of the 3D model.
● Preserves model integrity while hiding the encrypted data.
Implementation
Step 1: Importing the 3D Model

● Import the 3D model as the cover image.

CoverImage = Import["ExampleData/CThead.tiff", "Image3D"]

Step 2: Slicing the 3D Model

● Slice the 3D model into 2D segments.

ImageSlices = Image3DSlices[CoverImage]

ImageSlicesLength = Length[ImageSlices]
Step 3: Message selection and conversion
● Select and convert a message into ASCII codes, then into a bit stream.
TakenMessage = StringTake[ExampleData[{"Text", "Hamlet"}], 8192]
takenMessageAscii = ToCharacterCode[TakenMessage];
takenMessageToBits = IntegerDigits[takenMessageAscii, 2, 8] // Flatten // Normal;

Step 4: Generating secret key

● Generate a secret key using a chaotic function and thresholding.
sol = RecurrenceTable[{x[n + 1] == 3.6*Tan[x[n]]*(1 - x[n]), x[0] == 0.5}, x, {n, 1,
messageLength}]
key = ... // Generate key based on sol
Step 5: Encryption and Embedding

● Encrypt the message using the generated key.

● Embed the encrypted message into the 2D slices using Full LSB technique.

CodeEncrypted = ... // Encrypt message using key

Slices = ... // Embed CodeEncrypted into 2D slices

model = Image3D[Slices, "Byte"]

Step 6: Extraction and Decryption

● Extract and decrypt the embedded message from the stego 3D model.

modelSlices = Image3DSlices[model];

MessageBack = ... // Extract and decrypt the hidden message

Performance evaluation
● Performance Metrics Used:
● MSE, PSNR, SSIM, entropy, R measurement, IF, NCC, histogram analysis, image distance,
and image difference.
● Images Used:
● Two 3D images - CTEngine and CTHead from Wolfram Mathematica® library.
● Insights:
● Visual Analysis: Differences between cover and stego images are not easily detectable
through simple visual analysis.
● Importance of Metrics: The performance metrics are essential to evaluate the proposed
scheme accurately.
● Key Findings:
● Histogram Analysis: No visible differences between cover and stego images.
● MSE: Favorable results with low values (CTHead: 0.499938, CTEngine: 0.499101).
● PSNR: High quality indicated by values (CTHead: 51.1416, CTEngine: 51.1489).
● SSIM: Near-ideal values close to 1 (CTHead: 0.991605, CTEngine: 0.970397).
● IF, R Measurement, NCC: Values approaching ideal conditions.
● Entropy: Close similarity between cover and stego images (CTHead: same, CTEngine:
slightly different).
● Image Distance: Values indicating proximity (CTHead: 7.43859, CTEngine: 7.06277).
● Image Difference: Indicates minimal differences, almost canceling out elements.
Future scope
Enhanced Encryption Methods:
● Research and implement advanced encryption techniques like quantum or
homomorphic encryption for stronger protection against decryption attacks.

Machine Learning Integration:

● Explore the integration of AI algorithms to intelligently embed and detect
watermarks, improving adaptability to diverse data types and potential
vulnerabilities.
Conclusion
In summary, this thesis proposed a robust double-layered steganography scheme for
watermarking 3D models. Leveraging chaotic function encryption and LSB embedding,
the scheme effectively conceals sensitive information within 3D images while
maintaining visual integrity. The evaluation demonstrated promising results comparable
to existing schemes. This research opens avenues for heightened security in 3D data
and lays the groundwork for future advancements in secure data embedding within 3D
models.
THANK YOU!