Visual data security by data hiding and/or encryption William PUECH - - PowerPoint PPT Presentation

Visual data security by data hiding and/or encryption

William PUECH

ICAR (Image & Interaction) LIRMM, UMR 5506, CNRS - University of Montpellier

October 14, 2016

• W. Puech (ICAR)

Visual data security October 14, 2016 1 / 27

Visual data security

Visual data security: problem

Safe transfer and visualization on line in real time for low powered systems (wireless devices).

• W. Puech (ICAR)

Visual data security October 14, 2016 2 / 27

Visual data security

Visual data security: problem

Hybrid coding for safe transmission Encryption, data hiding and compression Images, image sequences, videos and 3D objects

• W. Puech (ICAR)

Visual data security October 14, 2016 3 / 27

Visual data security

Visual data security: problem

Hybrid coding for safe transmission Encryption, data hiding and compression Images, image sequences, videos and 3D objects Image compression

• W. Puech (ICAR)

Visual data security October 14, 2016 3 / 27

Visual data security

Visual data security: problem

Hybrid coding for safe transmission Encryption, data hiding and compression Images, image sequences, videos and 3D objects Image encryption

• W. Puech (ICAR)

Visual data security October 14, 2016 3 / 27

Visual data security

Visual data security: applications

• W. Puech (ICAR)

Visual data security October 14, 2016 4 / 27

Visual data security

Visual data security: solution

Hybrid coding for safe transmission Robust to noise Compatible with compression Fast: access in real time The secret is based on a key (secrete or private key)

The Algorithm is known Principle of Kerckhoffs [KER 83]

Norms and standards

• A. Kerckhoffs.

La cryptographie militaire. Journal des sciences militaires, vol. 9, pp. 5–38, 1883.

• W. Puech (ICAR)

Visual data security October 14, 2016 5 / 27

Visual data security

Visual data security: solution

Image data hiding The art to embed a message in a image [COX 08] :

invisibility: statistically invisible no removable: robust to transformations payload: size of the hidden message security: robust to attacks complexity: real time application

Data hiding: large payload Steganography: invisibility Watermarking: robust to attacks

• I. Cox.

Digital Watermarking and Steganography. The Morgan Kaufmann Series in Multimedia Information and Systems, M. Kaufmann, Ed. Morgan Kaufmann Publishers, 2008.

• W. Puech (ICAR)

Visual data security October 14, 2016 6 / 27

Visual data security

Visual data security: solution

Image encryption The art to mask the data:

confidentiality: data protection authentication: emitter and receiver integrity: ensure the totality and the content of the data non repudiation: ACK

For visual data:

Image encryption

example

Perceptual signature

example

• I. Cox.

Digital Watermarking and Steganography. The Morgan Kaufmann Series in Multimedia Information and Systems, M. Kaufmann, Ed. Morgan Kaufmann Publishers, 2008.

• W. Puech (ICAR)

Visual data security October 14, 2016 6 / 27

Visual data security

Visual data security: some results

Color protection by watermarking.

• M. Chaumont and W. Puech.

A 8-Bit-Grey-Level Image Embedding its 512 Color Palette. 16th European Signal Processing Conference, EUSIPCO’08 aug. 25-29, Lausanne, Switzerland.

• M. Chaumont, W. Puech and C. Lahanier.

Securing Color Information of an Image by Concealing the Color Palette. Journal of Systems and Software, Elsevier, 2012.

• W. Puech (ICAR)

Visual data security October 14, 2016 7 / 27

Visual data security

Visual data security: some results

Selective encryption of H264 videos (a) QP = 18, (b) QP = 24, (c) QP = 30 et (d) QP = 36.

J.M. Rodrigues, W. Puech and A. Bors. Selective Encryption of Human Skin in JPEG Images. Proc. International Conference on Image Processing IEEE ICIP-2006, pp. 1981-1984, Atlanta, US, october 2006.

• Z. Shahid, M. Chaumont and W. Puech.

Fast Protection of H.264/AVC by Selective Encryption of CAVLC and CABAC for I & P

• frames. IEEE Transactions on Circuits and Systems for Video Technology, 21(5)

:565-576, May 2011.

• W. Puech (ICAR)

Visual data security October 14, 2016 8 / 27

Visual data security

Visual data security: some results

Blink measure

BM =

• n−1
• i=1

m

• k=1

(Ioi,k − Ioi+1,k)2 −

n−1

• i=1

m

• k=1

(Iei,k − Iei+1,k)2

• (n−1)m
• L. Dubois, W. Puech and J. Blanc-Talon.

Reduced Selective Encryption of Intra and Inter Frames of H.264/AVC Using Psychovisual

• Metrics. Proc. International Conference on Image Processing IEEE ICIP-2012, Orlando,

US, october 2012.

• W. Puech (ICAR)

Visual data security October 14, 2016 9 / 27

Visual data security

Visual data security: some results

Scalable 3D visualization by watermarking.

• K. Hayat, W. Puech and G. Gesquière.

Scalable 3D Terrain Visualization through Reversible JPEG2000 Based Blind Data Hiding. IEEE Transactions on Multimedia, IEEE, vol. 10, n 7, pp. 1261-1276, Nov. 2008.

• K. Hayat, W. Puech, N. Islam and G. Gesquiere.

Seamless Heterogeneous Tessellation via DWT Domain Smoothing and Mosaicking. EURASIP Journal of Advances in Signal Processing Volume 2010, 15-pages.

• W. Puech (ICAR)

Visual data security October 14, 2016 10 / 27

Visual data security

Visual data security: some results

3D watermarking a) Original 3D object “Smilodon” 508796 vertices, b) 3D Object watermarked with 314071 bits (= 38.3 kbytes)

P . Amat, W. Puech, S. Druon and J.P . Pedeboy. Lossless 3D Steganography Based on MST and Connectivity Modification. Signal Processing: Image Communication, Elsevier, vol. 25, n◦ 6, pp. 400-412, July 2010.

• N. Tournier, W. Puech, G. Subsol and J.P

. Pedeboy. Sensitivity Analysis of Euclidean Minimum Spanning Tree for 3D Watermaking. Proc. SPIE, Electronic Imaging, 3D Image Processing, San-Francisco, CA, USA: SPIE, IS&T, January , 2012.

• W. Puech (ICAR)

Visual data security October 14, 2016 11 / 27

Visual data security

Visual data security: some results

• M. Pinto, W. Puech and G. Subsol.

Protection of JPEG compressed e-comics by selective encryption.

• Proc. International Conference on Image Processing IEEE ICIP-2013, Melbourne,

Australia, september 2013.

• W. Puech (ICAR)

Visual data security October 14, 2016 12 / 27

Visual data security

Visual data security: steganalysis

• L. Pibre, J. Pasquet, D. Ienco, and Marc Chaumont.

Deep learning is a good steganalysis tool when embedding key is reused for different images, even if there is a cover source-mismatch

• Proc. Electronic Imaging, Media Watermarking, Security, and Forensics,

San-Francisco, CA, USA: IS&T, 2016.

• H. Abdulrahman, M. Chaumont, P

. Montesinos, and B. Magnier. Color Image Steganalysis Based on Steerable Gaussian Filters Bank

• Proc. ACM IH&MMSec’2016, Vigo, Galicia, Spain, June 20-22, 2016.
• W. Puech (ICAR)

Visual data security October 14, 2016 13 / 27

Visual data security

Data hiding in encrypted images

Coding step First step: encryption Second step: data hiding

• W. Puech (ICAR)

Visual data security October 14, 2016 14 / 27

Visual data security

Data hiding in encrypted images

Decoding step 3 possible scenarios Message extraction then decryption

• W. Puech (ICAR)

Visual data security October 14, 2016 15 / 27

Visual data security

Data hiding in encrypted images

Decoding step 3 possible scenarios Message extraction and decryption in parallel

• W. Puech (ICAR)

Visual data security October 14, 2016 15 / 27

Visual data security

Data hiding in encrypted images

Decoding step 3 possible scenarios Decryption then message extraction

• W. Puech (ICAR)

Visual data security October 14, 2016 15 / 27

Visual data security

Data hiding in encrypted images: previous work

Original image Secret key k image Encrypted Encryption Data−hiding image encrypted Marked message Marked encrypted image Secret key k Message removing Decryption and Original image Extraction of hidden message message

(a) (b) Noise removing in the encrypted domain. a) Coding, b) Decoding.

• W. Puech, M. Chaumont, and O. Strauss.

A Reversible Data Hiding Method for Encrypted Images. Proc. SPIE, Electronic Imaging, Security, Forensics, Steganography, and Watermarking of Multimedia Contents X, volume 6819, pages 68191E-1-68191E-9, San Jose, CA, USA, January 2008.

• N. Islam and W. Puech.

Noise Removing in Encrypted Color Image by Statistical Analysis. Proc. SPIE, Electronic Imaging, MediaWatermarking, Security and Forensics, San-Francisco, CA, USA: SPIE, IS&T, January , 2012.

• W. Puech (ICAR)

Visual data security October 14, 2016 16 / 27

Visual data security

Data hiding in encrypted images: previous work

• W. Puech, M. Chaumont, and O. Strauss.

A Reversible Data Hiding Method for Encrypted Images. Proc. SPIE, Electronic Imaging, Security, Forensics, Steganography, and Watermarking of Multimedia Contents X, volume 6819, pages 68191E-1-68191E-9, San Jose, CA, USA, January 2008.

• W. Puech (ICAR)

Visual data security October 14, 2016 17 / 27

Visual data security

Entropy analysis of an image encrypted block (1/2)

The information content of a message S can be measured with the entropy H(S). If a message S has M symbols α, with 0 ≤  < M, and the probability of symbol α is P(α), the entropy H(S) (in bits/symbol) is defined as: H(S) = −

M−1

• =0

P(αj)log2(P(αj)). (1) Definition (Maximum entropy) An encryption function Ek() is efficient if and only if the entropy of the encrypted message is maximum: H(Y) = H(Ek(X)) = b bits/symbol ∀X, ∀k ∈ K. Theorem (Relation between entropies) If an encryption function Ek() is efficient, then the entropy H(Y) of an encrypted message Y is greater than the entropy H(X) of the

• riginal message X:

H(Y) = H(Ek(X)) ≥ H(X). (2) Proof. From Definition, since H(Y) = b bits/pixel, where b is the maximum value, thus H(X) ≤ b, then H(Y) ≥ H(X).

• W. Puech (ICAR)

Visual data security October 14, 2016 18 / 27

Visual data security

Entropy analysis of an image encrypted block (2/2)

Scientific challenges What is the minimal size of a block so that the entropy of this block is significant? Trade-off between the payload, the quality of the decrypted image and the bit error rate of the extracted data? Possible leads Analyze the existing, Use statistical tests of hypotheses, Distribution study, Proposition of new data hiding scheme in encrypted images (based on homomorphic functions: E(mx ⊕ my) = E(mx) ⊗ E(my))

• W. Puech (ICAR)

Visual data security October 14, 2016 19 / 27

Visual data security

Visual data security: work in progress

New synchronization methods of vertex cloud for data hiding

PhD thesis of Vincent Itier, STRATEGIES Company

Generation and analysis of textured patterns to identify printed documents

PhD thesis of Iuliia Tkachenko, AUTHENTICATION INDUSTRIE Company

Authentication of acquisition devices of medical images by noise analysis contained in images

PhD thesis of Anas Kharboutly, Montpellier Hospital and IMAIOS Company

Image processing in the encrypted domain Confidentiality metrics - with temporal aspects Applications: remote surveillance, security, protection, culturage heritage, medical imaging, cell phone applications.

• W. Puech (ICAR)

Visual data security October 14, 2016 20 / 27

Visual data security

Synchronization of 3D vertices for data hiding. Joint embedding to the synchronization = ⇒ = ⇒

Mesh and first vertex v0 (1) Embedding on ρi+1 (2) Watermarked Hamil- tonian path (1): research of the nearest vertex (2): message embedding

• V. Itier, W. Puech, and J.-P

. Pedeboy. High capacity data-hiding for 3D meshes based on static arithmetic coding. In IEEE International Conference on Image Processing, 2015.

• V. Itier, N. Tournier, W. Puech, G. Subsol, and J.-P

. Pedeboy Analysis of an EMST-based path for 3D meshes. Computer-Aided Design, Elsevier, vol. 64, p : 22-32, 2015

• W. Puech (ICAR)

Visual data security October 14, 2016 21 / 27

Visual data security

Graphical codes for authentication of printed documents

Two storage levels First level for standard application Second level with a q-aire alphabet Increasing of the capacity Sensitive to the copy

• Iu. Tkachenko, W. Puech, C. Destruel, O. Strauss, J.-M. Gaudin and C. Guichard

Two-Level QR Code for Private Message Sharing and Document Authentication . IEEE Transactions on Information Forensics and Security , vol. 11, n◦ 3, p : 571-583, March 2016

• W. Puech (ICAR)

Visual data security October 14, 2016 22 / 27

Visual data security

Identification of medical devices by noise analysis in images (a) (b) a) CT-Scanner Identification b) Identification results.

• A. Kharboutly, W. Puech, G. Subsol and D. Hoa.

IMPROVING SENSOR NOISE ANALYSIS FOR CT-SCANNER IDENTIFICATION. 23rd European Signal Processing Conference (EUSIPCO), Nice, France, September, 2015.

• W. Puech (ICAR)

Visual data security October 14, 2016 23 / 27

Visual data security

GDR ISIS : http://gdr-isis.fr/ DSA de l’axe A du thème D "Compression et Protection" Responsable de l’Action "Protection Multimédia"

2016 : Journée Saillance visuelle et applications au tatouage et à la compression d’images et de vidéos 2015 : Journée Biométrie, Indexation multimédia et Vie privée 2013 : Journée Extraction de preuves multimédia: détection de manipulations - identification et authentification de contenus ou de personnes 2013 : Journée commune TRECVID + Qualité et Protection 2012 : Journée plénière compression et protection 2012 : Journée Interactions entre cryptographie et dissimulation d’information (stéganographie, tatouage, fingerprinting) 2011 : Journée Fingerprint - Action 2 Protection des données visuelles

• W. Puech (ICAR)

Visual data security October 14, 2016 24 / 27

Visual data security

Visual data security: quality metric

Fouille d’images dans le domaine chiffré :

Indexation, Classification, Recherche.

Métriques de qualité :

Avec ou sans référence, Métrique de confidentialité, Prise en compte de l’aspect temporel.

• W. Puech (ICAR)

Visual data security October 14, 2016 25 / 27

Visual data security

Visual data security: quality metric

• W. Puech (ICAR)

Visual data security October 14, 2016 25 / 27

Visual data security

Visual data security: quality metric

• W. Puech (ICAR)

Visual data security October 14, 2016 25 / 27

Visual data security

Visual data security: quality metric

• W. Puech (ICAR)

Visual data security October 14, 2016 25 / 27

Visual data security

Visual data security: quality metric

• W. Puech (ICAR)

Visual data security October 14, 2016 25 / 27

Visual data security

Visual data security: quality metric

• W. Puech (ICAR)

Visual data security October 14, 2016 25 / 27

Visual data security

Visual data security: quality metric

• W. Puech (ICAR)

Visual data security October 14, 2016 25 / 27

Visual data security

Perceptual signatures: data integrity

Signature of a text

M1 = “Aujourd’hui il fait beau dans le sud de la France, même si il y a un peu de vent...” S1 = Ox2534A8C08E12F4A8 M2 = “Aujourd’hui il fait beau dans le sud de la France, même si il y a un peu de mistral...” S2 = Ox3D68AB9310E38B51

Signature of an image

S1(original image (760 kB)) = S2(compressed image (224 kB))

back

• W. Puech (ICAR)

Visual data security October 14, 2016 26 / 27

Visual data security

Perceptual signatures: data integrity

Signature of a text

M1 = “Aujourd’hui il fait beau dans le sud de la France, même si il y a un peu de vent...” S1 = Ox2534A8C08E12F4A8 M2 = “Aujourd’hui il fait beau dans le sud de la France, même si il y a un peu de mistral...” S2 = Ox3D68AB9310E38B51

Signature of an image

S1(original image (760 kB)) = S2(compressed image (224 kB))

back

• W. Puech (ICAR)

Visual data security October 14, 2016 26 / 27

Visual data security

Image encryption

(a) (b) (c) (d) a) Original image, b) histogram, c) encrypted image by scrambling, d) histogram of the encrypted image.

• W. Puech (ICAR)

Visual data security October 14, 2016 27 / 27

Visual data security

Image encryption

(a) (b) (c) (d) a) Original image, b) histogram, c) encrypted image with a stream cipher algorithm, d) histogram of the encrypted image. back

• W. Puech (ICAR)

Visual data security October 14, 2016 27 / 27