Lecture 13 Steganalysis and Watermarking

Steganalysis

• Definition: Process of detecting secret messages within steganographic methods.either by simply detecting the presence of a secret message or by extracting and potentially destroying it (Andriotis et al., 2013)

  – Steganalysis is sometimes simply called attacks

  – Similar concept to cryptanalysis for cryptology

• Goal: to detect the existence of hidden communications (i.e., the existence of a secret message) and ideally but NOT necessarily extracting the secret message

Steganography defeat Techniques

• Some ways to remove a payload (secret message)
  when the steganography method used is not sophisticated

  like LSB (Least Significant Bit)

  • 1.Compression,

  • 2.image resizing,

  • 3. color palette change,

  • 4. luminescence change,

  • 5. filtering,

  • 6. Random noise insertion,

  • 7. property modification.

    Most steganographic methods leaves traces in the

    stego media

    – (a) Therefore, changing their statistical properties

    – (b ) A lot of steganalysis techniques are based in the detection

    of these statistical changes

    – (c ) Steganography tools might also add particular traces or

    patterns/signatures

    Steganalysis detection Techniques

  • Detection: Statistical analysis to find anomalies e.g., entropy increase.

  • Statistical analysis [1]

    – Uses statistical properties & filtering to look for statistical

    anomalies according to a norm

    – E.g., when a carrier contains a payload (e.g., added via LSB), the

    randomness of the carrier bits changes, and the difference can

    be detected

    • This means that the entropy (randomness) usually gets higher

• Steganalysis Tools/Algorithms:

• When embedding secret information into a carrier/cover

  media, usually:
  Detect entropy increase, 2. decrease in color/DCT (Discrete Cosine Transform) coefficient frequency differences, 3. iNCREASE in similar adjacent colors.

• Steganalysis Challenges:

• Challenges in detecting deviations from the norm in some statistical properties:

  1. high number of False positives,

  2. norm definition very complex,

  3.It is possible, after the hiding process, to use additional

  techniques to make the result closer to any norm (Making detection more difficult)

  4. payload size dependency (smaller payloads are harder to detect)

  5. very media dependency.

• Steganalysis Classification

  • 1. Anomaly-based: Histogram & statistical analysis, file property changes, visual/audible changes/distortions.

  • 2. Signature-based: Patterns/traces/signatures consistent with the of specific tools used

• Steganalysis Tools:

• Steganalysis tools are able to identify different

  steganography schemes

  – E.g., LSB, Invisible Secrets, Camouflage, appendX, Fuse, StegoExpose, Stegodetect

• Steganalysis meets Cryptanalysis:

• Confidentiality of the secret message may be enforced when a stego key is used to protect it

  – This aids in hiding the secret, but isn’t encreyption

  • The secret message can be further protected with standard encryption of the raw payload prior to embedding

  • Revealing the secret message may involve

  – Brute force to get the password

  – Cracking the stego key and bypass encryption

Watermarking

• Definition: Protecting intellectual property via embedded marks in media files.

  • Used for ownership/authenticity proof, copyright protection, content authentication.

    – The mark can be used to proof ownership or authenticity

    – Crucial mechanism used for copyright protection and

    content authentication

• Watermarking Types:

  • Visible

  • Invisible

• Watermarking Applications:

  • Source of leaked movie identification

  • Digital rights management (DRM)

  • Broadcast monitoring.

• Watermarking Modules:

• Watermarking is intended to be detectable under certain circumstances, either by humans or by algorithms (i.e., invisible to humans).

• Watermarking Embedding module → Watermark detection & decoding modules.

• Robustness a major requirement:

• Classification of watermarking techniques according to the level of robustness offered

  • 1. Robust: watermark is designed to resist host signal manipulations; usually employed in Intellectual Property Rights (IPR) protection applications

  • 2. Fragile: Vulnerable to modification; easy to implement., used for authentication by user

  • 3. Semi-fragile: designed to be Robust against selected manipulations.

• Robustness Attacks:

  1. Geometric attacks aim to destroy the synchronization between a watermark and a detector.” (destroy synchronization /Removal (eliminate watermark).

  2. Removal attacks aim to eliminate the watermark without knowledge of the watermarking algorithm and key.”

• Characteristics:

• 1. Watermark should cause minimal impact on the original media, i.e., minimal degradation

• 2. The watermark should be related to the original media, i.e., related to its ownership, its distribution, its use

• Techniques:

• Similar to techniques used in steganography, using

  – Spatial domain: matrix of pixels of a media file

  – Frequency domain: frequency spectrum/coefficients of a media file via some sort of transformation, e.g.,

  • Discrete Fourier Transform (DFT)

  • Discrete Cosine Transform (DCT)

  • Wavelet Transform (WT)

Watermarking vs. Steganography

• Similarities: Both classified under data hiding/multimedia forensics and use common techniques. They are distinct but related disciplines in the field of digital image processing

  – Often classified under “data hiding” and “multimedia forensics”

  – Steganography is also considered as an “anti-forensics” approach

  – They use common techniques and may be subject to similar attacks

  BUT there are Differences:

  • Watermarking vs Steganography – Comparison Table

    Feature	Watermarking	Steganography
    Detection	Intended to be detectable	Intended to be undetectable
    Purpose	Protect intellectual property	Conceal information
    Alteration	Minimal impact on the original image	Subtle changes to hide data
    Visibility	Often visible	Hidden and imperceptible