Information Hiding: Steganography and Watermarking

Dr. Mohammed Al-Mualla and Prof. Hussain Al-Ahmad

Multimedia Communication and Signal Processing (MCSP) Research Group Etisalat College of Engineering
P.O.Box: 980, Sharjah, UAE Tel: +971 6 5611333, Fax: +971 6 5611789, e-mail: {almualla, alahmad}@ece.ac.ae

Outline
Introduction to Information Hiding Steganography
Definition and History Applications Basic Principles Examples of Techniques Demos Definition and History Applications Basic Principles Requirements Attacks Evaluation and Benchmarking Examples

Watermarking

Information Hiding
Information Hiding is a general term encompassing many subdisciplines Two important subdisciplines are: steganography and watermarking Steganography:
Hiding: keeping the existence of the information secret

Watermarking:
Hiding: making the information imperceptible

Information hiding is different than cryptography (cryptography is about protecting the content of messages)

Steganography: Definition and History

Steganography: derived from the Greek words steganos which means covered and graphia which means writing, i.e. covered writing It is the art of concealed communication; the very existence of a message is secret Examples of old steganography techniques:
Writing on shaved heads Invisible ink Microscopic images

Steganography: Applications
Unobtrusive communication
Military and intelligence agencies Criminals !! Normal people

Plausible deniability
Fair voting Personal privacy Limitation of liability

Anonymous communication
Vote privately Make political claims Access censored material Preserve free speech

Steganography: Basic Principles

TRANSMITTER cover c RECEIVER

stego-object stego-object s s Encoder Insecure Decoder (Hiding Mechanism) (Extraction Mechanism) Channel

message m

key k

key k

message m

Steganography: Examples of Techniques

1. Substitution Techniques: Substitute redundant parts of a cover with a secret message Example: Least Significant Bit (LSB) Substitution
Choose a subset of cover elements and substitute least significant bit(s) of each element by message bit(s) Message may be encrypted or compressed before hiding A pseudorandom number generator may be used to spread the secret message over the cover in a random manner Easy but vulnerable to corruption due to small changes in carrier

Steganography: Examples of Techniques

2. Transform Domain Techniques: Embed secret message in a transform space (e.g. frequency domain) of cover Example: Steganography in the Discrete Cosine Transform (DCT) domain
Split the cover image into 88 blocks. Each block is used to encode one message bit Blocks are chosen in a pseudorandom manner The relative size of two pre-defined DCT coefficients is modulated using the message bit The two coefficients are chosen from middle frequencies (trade off between robustness and imperceptibility)

Steganography: Examples of Techniques

3. Spread Spectrum Techniques:
Adopt ideas from spread spectrum communication where a signal is transmitted in a bandwidth in excess of the minimum necessary to send the information In other words, the message is spread over a wide frequency bandwidth The SNR in every frequency band is small (difficult to detect) Even if parts of the message are removed from several bands, enough information is present in other bands to recover the message Thus, it is difficult to remove the message completely without entirely destroying the cover (robustness)

Steganography: Examples of Techniques

4. Statistical Techniques: Encode information by changing several statistical properties of a cover The cover is split into blocks. Each block is used to hide one message bit If the message bit is 1 then the cover block is modified, otherwise the cover block is not modified Difficult to apply in many cases, since a good test must be found which allows distinction between modified and unmodified cover blocks

Steganography: Examples of Techniques

5. Distortion Techniques: Store information by signal distortion The encoder applies a sequence of modifications to the cover. This sequence corresponds to the secret message The decoder measures the differences between the original cover and the distorted cover to detect the sequence of modifications and consequently recover the secret message Not useful in many applications since the decoder must have access to the original cover Example: vary the distance between consecutive lines or words to transmit secret information

Steganography: Examples of Techniques

6. Cover Generation Techniques: Encode information in the way a cover is generated Example: Automated Generation of English Text
Use a large dictionary of words categorised by different types, and a style source which describes how words of different types can be used to form a meaningful sentence Transform message bits into sentences by selecting words out of the dictionary which conforms to a sentence structure given in the style source

Steganography: Demo 1
Information Hidden in an Audio File

cover c

stego-object stego-object s s Encoder Insecure Decoder (Hiding Mechanism) (Extraction Mechanism) Channel

message m

message m

Steganography: Demo 1
Information Hidden in an Audio File
Cover is a CD quality audio file (sampling frequency: 44.1 kHz, resolution: 16 bits, stereo) Message is a 256256 24-bit colour image
12 samples of Left audio channel 1 pixel of image LSB Substitution 16 bits

24 bits

12 samples of Right audio channel

Steganography: Demo 2
Information Hidden in an Image File

cover c stego-object stego-object s s Encoder Insecure Decoder (Hiding Mechanism) (Extraction Mechanism) Channel

message m

message m

Steganography: Demo 2
Information Hidden in an Image File
Cover is a 512512 8-bit monochrome image Message is an 8 kHz, 8-bits, mono audio file

1 sample of audio LSB Substitution

8 pixels of image

8 bits

8 bits

Watermarking: Definition
Watermarking: is the practice of imperceptibly altering a cover to embed a message about that cover Watermarking is closely related to steganography, but there are differences between the two:
In watermarking the message is related to the cover Steganography typically relates to covert point-to-point communication between two parties Therefore, steganography requires only limited robustness Watermarking is often used whenever the cover is available to parties who know the existence of the hidden data and may have an interest in removing it Therefore, watermarking has the additional notion of resilience against attempts to remove the hidden data

Watermarks are inseparable from the cover in which they are embedded. Unlike cryptography, watermarks can protect content even after they are decoded

Watermarking: History
More than 700 years ago, watermarks were used in Italy to indicate the paper brand and the mill that produced it By the 18th century watermarks began to be used as anticounterfeiting measures on money and other documents The term watermark was introduced near the end of the 18th century. It was probably given because the marks resemble the effects of water on paper The first example of a technology similar to digital watermarking is a patent filed in 1954 by Emil Hembrooke for identifying music works In 1988, Komatsu and Tominaga appear to be the first to use the term digital watermarking About 1995, interest in digital watermarking began to mushroom

Watermarking: Applications
1. Copyright protection
Most prominent application Embed information about the owner to prevent others from claiming copyright Require very high level of robustness

2. Copy protection
Embed watermark to disallow unauthorized copying of the cover For example, a compliant DVD player will not playback or copy data that carry a copy never watermark

3. Content Authentication
Embed a watermark to detect modifications to the cover The watermark in this case has low robustness, fragile

Watermarking: Applications
4. Transaction Tracking
Embed a watermark to convey information about the legal recipient of the cover This is useful to monitor or trace back illegally produced copies of the cover This is usually referred to as fingerprinting

5. Broadcast Monitoring
Embed a watermark in the cover and use automatic monitoring to verify whether cover was broadcasted as agreed

Watermarking: Basic Principles

TRANSMITTER cover c watermarked data watermarked data s S Insecure Channel (Attacks) RECEIVER watermark w or cover c

Encoder
(Embedding)

Decoder
(Extraction)

watermark w

key k

key k

watermark w or confidence measure

Watermarking: Requirements
Imperceptibility
The modifications caused by watermark embedding should be below the perceptible threshold

Robustness
The ability of the watermark to resist distortion introduced by standard or malicious data processing

Security
A watermark is secure if knowing the algorithms for embedding and extracting does not help unauthorised party to detect or remove the watermark

Watermarking: Requirements
Payload
The amount of information that can be stored in a watermark

Informed (nonoblivious, or private) Watermarking:

The original unwatermarked cover is required to perform the extraction process

Blind (oblivious, or public) Watermarking:

The original unwatermarked cover is NOT required to perform the extraction process

Watermarking: Attacks
Signal enhancement (sharpening, contrast enhancement, etc.) Additive and multiplicative noise (gaussian, uniform, etc.) Filtering (High pass, low pass, linear, nonlinear, etc.) Lossy compression (JPEG, MPEG-x, H.26x, etc.) Geometric transforms (translation, rotation, etc.) Data reduction (cropping, clipping, etc.) Transcoding (MPEG2 H.263, etc.) D/A and A/D conversion (print-scan, etc.) Multiple watermarking Collusion attack Mosaic attack

Evaluation and Benchmarking

Subjective Evaluation:
Subject is asked to rate the perceptibility of the artefacts (5=imperceptible, 4=perceptible but not annoying, 3=slightly annoying, 2=annoying, 1=very annoying)

Watermarking:

Objective Evaluation:
MSE, SNR, PSNR, NCC, etc.

Benchmarking Software:
Unzign StirMark CheckMark OptiMark

Watermarking: Examples
Informed Watermarking Method of Cox et al.
Original image D Extract largest N coefficients

DCT

V = [v1 , v2 , K, v N ]

key k

Random Number Generator

X = [x1 , x2 , L, x N ]

Embed

vi = vi (1 + xi )
V

Watermarked Image D

IDCT

Insert back N modified coefficients

ENCODER

Watermarking: Examples
Informed Watermarking Method of Cox et al.
Original image D Attacked image D* DCT Extract largest N coefficients Extract N coeff. from same locations

V = [v1 , v2 , K, v N ]
Extract

DCT

V* = [v *1 , v *2 , K , v * N ]

xi * = (

vi * 1) / vi
X*

key k

Random Number Generator

X = [x1 , x2 , L, x N ]

Calculate similarity measure Decision

DECODER

Watermarking: Examples

original

= 0.1

= 0.4

= 0.6

original

N = 2000

N = 4000

N = 6000

Watermarking: Examples

original

JPEG q=100 NC=0.999

JPEG q=80 NC=0.993

JPEG q=10 NC=0.557

original

Watermarked NC=0.999

Cropped NC=0.672

Watermarking: Examples
Original Image 512 X 512 DCT Watermarked Image

IDCT Watermark 128 X 64 Watermarked Image

DCT-domain Embedding without Walsh Coding

Original Image 512 X 512

DCT IDCT

Walsh Coding

DCT-domain Embedding with Walsh Coding

128 x 64

Watermarking: Examples

Questions