Digital Watermarking for

Multimedia Security
R. Chandramouli
MSyNC:Multimedia Systems,
Networking, and Communications Lab
Stevens Institute of Technology
 Example:LSB Encoding

Original Image Watermark

Replace 4 and 7 LSBs of Original

4 LSBs Watermarked 7 LSBs Watermarked

 What is a Watermark?
 A watermark is a “secret message” that is embedded
into a “cover (original or host) message”.
 Only the knowledge of a secret key allows us to
extract the watermark from the cover message.
 Effectiveness of a watermarking algorithm is a
function of its
» Resilience to attacks.
» Capacity.
» Stealth.
 What Can Be Watermarked?
 Multimedia data.
» Video.
» Audio.
» Still Images.
» Documents.
 Software.
 Hardware designs.
 Multimedia Watermarks
 A digital watermark is a “secret key
dependent” signal “inserted” into digital
multimedia data.
 Watermark can be later detected/extracted in
order to make an assertion about the data.
 A digital watermark can be.
» Visible (perceptible).
» Invisible (imperceptible).
 Watermark Applications
 Proof of ownership.
» Prove ownership in a court of law.
» Simple copyright notice may not help for digital
multimedia data. Easily destroyed.
 Copy prevention or control.
» Recording device may inhibit recording a signal if
detected watermark indicates that it is prohibited.
» DVD video and digital music distribution.
 Content protection (visible watermarks).
» Destruction of watermark destroys the content.
 Watermark Applications
 Authentication.
» Detect if image/video has been altered.
» Digital cameras.
 Media Bridging.
» Bridge media such as magazines and the Internet.
» Digimarc.
 Broadcast Monitoring.
» Keep track of when and where an advertisement is
played.
» ConfirMedia from Verance.
 Watermark Applications

 Fingerprinting.
» Identify the source of an illegal copy.
» Unique watermark embedded in each copy.
» DiVX, a modified version of DVD.
 Secret Communications.
» Hide information such that general public do not
know its presence.
» Bin Laden hides attack plans in images on the
web – USA Today, Feb. 26, 2001.
 Requirements
 Requirements vary with application. For
example:
» Perceptually transparent - must not perceptually
degrade original content.
» Robust - survive accidental or malicious attempts
at removal.
» Oblivious or Non-oblivious - Recoverable with or
without access to original.
» Capacity – Number of watermark bits embedded.
» Efficient encoding and/or decoding.
Contradicting Requirements

Perceptual Transparency

Payload Robustness Security

Oblivious vs. Non-Oblivious

Example: Additive Watermarks

I(x,y) IW(x,y)

Multiply by gain
k factor k IW(x,y)= I(x,y)+ k  W(x,y)

RI 'W ( x , y )W ( x , y )  T 
W(x,y) detected
T 
W(x,y): Pseudo Random Pattern {-1,0,1} No W(x,y) detected
Additive watermarks in transform
domain: Spread Spectrum

fv

fh
 Watermark Attacks
 Active Attacks.
» Hacker attempts to remove or destroy the
watermark.
» Watermark detector unable to detect watermark.
» Key issue in proof of ownership, fingerprinting,
copy control.
» Not serious for authentication or covert
communication.
 Watermark Attacks
 Passive Attacks.
» Hacker tries to find if a watermark is present.
» Removal of watermark is not an aim.
» Serious for covert communications.
 Collusion Attacks.
» Hacker uses several copies of watermarked data
(images, video etc.) to construct a copy with no
watermark.
» Uses several copies to find the watermark.
» Serious for fingerprinting applications.
 Watermark Attacks
 Forgery Attacks.
» Hacker tries to embed a valid watermark.
» Serious in authentication.
» If hacker embeds a valid authentication
watermark, watermark detector can accept bogus
or modified media.
Example: Forging Photo I.D.
Millionaire with
Swiss bank
Bad
account
Forgery!

Thanks
Hong Man,
for this
picture!

Assistant
Professor
Content-based Watermarking
Original Watermark
JPEG Compression Attack
 Watermarking Research

 Information Theoretic Issues.

» Compute “watermarking capacity”
» How many watermark bits can be hidden in a
watermark channel ?
» Capacity when multiple watermarks are used.
 Decision Theoretic Issues.
» Robust watermark detector.
» Quickest possible watermark detection.
» Invariant to statistical uncertainties.
» Achieves desired false alarm and detection
probabilities.
 Watermarking Research
 Signal Processing Issues.
» What transforms to use ? – wavelet, DCT etc.
» Geometric invariant transform – scale, translation,
and rotation invariance.
 Watermarking protocols and system issues.
» Protocol used by sender and receiver for secure
transmission.
 Steganalysis.
» How to analyze multimedia data for the presence
of watermark(s).
» Secure Digital Music Initiative (SDMI) has been
hacked!
 Robust Watermarking Techniques
- A Communications Framework

Media Carrier Media Carrier

Coding and Decoding and

Water-
X +
Modulation Demodulation Watermark
mark

HVS
Noise/Attack Channel
Techniques Developed at
MSyNC Lab
(some results are in collaboration with
Iowa State U. and Polytechnic U.)
 Information Theoretic
Formulation
 A new watermark channel model.
» Multiplicative and additive attacks.
» Models geometric attacks (“killer” attack).
 Attack parameters assumed imperfectly
known.
 Lower and upper bounds on watermarking
capacity.
 Many current capacity estimates are specific
cases of proposed estimates.
 Capacity in the presence of multiple
watermarks using multiuser information
theory.
 Multiple Description
Watermarking
 A new framework for watermarking using
multiple description codes.
» Partition the source into multiple descriptions.
» Each description carries enough information.
» Allocate watermarking bits to descriptions subject
to distortion constraints.
» Experiments for secure wireless media
transmission is promising.
 Rate-distortion theory formulation.
Decision Theoretic Formulation
 How many pixels to watermark ?
» A new measure for watermarking capacity.
» Different from standard information theoretic
formulation.
» Perhaps a more practical measure.
 Fast watermark detector.
» Computation reduction.
» Useful for hardware implementation.
 Steganalysis
 A distributed detection framework for
analyzing images for presence of
watermarks.
 Theoretical formulation and practical
algorithm for steganalysis of LSB based
image watermarking.
 A new watermarking capacity measure called
“steganographic capacity”.
 Projected Applications of
Proposed Techniques

 Secure media streaming.

 Wireless multimedia security.
 Watermark error resilience coding.
 Error concealment for wireless multimedia
transmission.
 Refer to
» http://www.ece.stevens-tech.edu/~mouli/msync/index.html
 Million Dollar Question

Watermarking or Encryption or Combination?