Download presentation

Presentation is loading. Please wait.

Published byKadin Tow Modified about 1 year ago

1
Authentication of Paper Printed Documents Using Paper Characteristics Matúš Mihaľák ETH Z ürich joint work with Ivan Kočiš, Infotrust Slovakia

2
Vabo SPI '03 Brno 2 Introduction Typical Authentication: Stamps, seals, signatures, watermarks, holograms, etc., ? = ? = New technology -> better systems against forgery New technology -> better possibilities to counterfeit

3
Vabo SPI '03 Brno 3 Inspiration in Digital Documents’ Techniques Generally: Document is signed using public key cryptography. Signature is somehow attached to the paper document. Drawback: Photocopy of such a document is a valid document InfoMark technology:

4
Vabo SPI '03 Brno 4 ID of a paper Scanned Paper Image size is a problem... Need image features: FINGERPRINT of image f

5
Vabo SPI '03 Brno 5 Paper statistics... Histogram d Pixel correlation Mutual Intensity Occurance Dist=1, 3 and 5

6
Vabo SPI '03 Brno 6 Local extremes (i,j) - local extreme iff (i,j) - global extreme on subimage (i-R …i+R, j-R …j+R) 2R (i,j) R R – parameter FINGERPRINT – set E of local extremes

7
Vabo SPI '03 Brno 7 Moments Image f – probability density function Statistical characteristics: Moments m ks = i j i k j s f(i,j) FINGERPRINT – first N moments from every square a b

8
Vabo SPI '03 Brno 8 Fourier coefficients Frequency domain of an image f Discrete Fourier Transform: F(u,v)= k l e -2 i(uk/M+vl/N). f(k,l) FINGERPRINT – first K coefficients of Fourier transform from every square

9
Vabo SPI '03 Brno 9 Fingerprint similarity measurement Given 2 images f 1 and f 2 Local extremes – E 1 and E 2 |{E 1 E 2 }| / |{E 1 E 2 }| - occurence ratio Moments and Fourier coeffs – x and y Correlation coefficient x and y – variances of x and y

10
Vabo SPI '03 Brno 10 Authentication Scheme 1. Scanning of paper in transparency mode 2. Feature extraction from image Signature: 3. Digital signature of features and document 4. Printing signature and document using InfoMark Verification: 3. Reading paper features from InfoMark 4. Comparing features and document content

11
Vabo SPI '03 Brno 11 Results - Local Extremes F SizeDifferenceLow matchHigh FailRIP %49.47%18.33%81 vs %50.47%9.57%161 vs % 55.34%11.22%167 vs %68.41%18.11%87 vs %75.16%10.18%16Gauss %77.39%16.92%8Gauss 5 Extremes may differ by 3 in coordinates

12
Vabo SPI '03 Brno 12 Results - Moments a x bF SizeDifferenceLow matchHigh FailNIP 32 x vs x vs x x G G 5 32 x x vs vs. 1

13
Vabo SPI '03 Brno 13 Results – Fourier descriptors F SizeDifference|Low Match||High Fail|a x bN x x x x x x 322

14
Vabo SPI '03 Brno 14 The End Thank you for your attention Questions?

15
Vabo SPI '03 Brno 15 Security – Local extremes Image of the size 256 x 256 R = 8, #extremes = 370 => P[(i,j) is extreme]< Benevolence ± 2 in coordinates => P[ex]< P[>60% match] = P[61]+P[62]+..+P[100] P[k] = (370 choose k) * P[ex] k P[> 60%] < x

Similar presentations

© 2016 SlidePlayer.com Inc.

All rights reserved.

Ads by Google