10 th International Conference of Information System and

10 th International Conference of Information System and Security 16 -20 Dec 2014, IDRBT, Hyderabad Ph D Symposium Automatic Authentication of Printed Security Documents By Biswajit Halder Dept. Of Computer Science University Of Burdhwan, W. Bengal, India biswajithalder 88@gmail. com

Security Documents Security Document achieved by 4 -Layer Techniques – using stock of paper, special type of ink or coating, unique no of barcode and complex graphics respectively. High Security Documents-Currency Note -deeds -Wills -passports Medium Security documents-shares and bonds -Checks Low security Documents -Lottery Tickets -Drafts

Security Category Low Overt Medium High -See-through register -Shadow image -Unique identifier -Watermark -Bearer ’s signature -Digital facial image -See-through register -Shadow image -Unique identifier -Watermark -Diffractive optically -Bearer’s signature -Digital facial image -See-through register -Shadow image -Unique identifier -Watermark -High resolution Semi- -High resolution printing processes Covert -Security-type printing processes features Covert -Security fibre paper -Screen angle modulation -Security-type printing features -Security fibre paper -Hidden image -Screen angle modulation -Security ink -High resolution printing processes -Security-type printing features -Security fibre paper -Hidden image -Screen angle modulation -Security ink -Security threads

Problem and Demand Problem – • With advent of high quality scanning and printing technology, generation of fake documents is an easy affair. • Amount of fake documents is now a serious threat to our society. • To check every document in question, asking help of forensic experts is an unrealistic solution, • Forensic experts may take time to give their views. • The frequency of such documents in question may be large Demand is – - A Quick Answer - Ability to process large number of Document - Sometimes low cost solution

Present Area of Authentication Checking Generally Document Examination done on following areas – -Forgery Specialists - Analysis by public or private experts -Document examiner -Analysis through laboratory equipment - Checking Document Dating - verification of age - Fraud investigator - focuses on money trail and criminal intent - Testing Paper – Generally tested by chemical methods - Testing Ink - Generally tested by chemical methods - Handwriting Analysts - Typewriting Analysts

Why Automation? • Speed and Accuracy check through maximum no. of features • More Useful in Electronic check image for bankcheck Processing • Automatic banknote recognition with features level analysis. • Applications in Automated Teller Machines (ATM) • Stresses to be “self-defending”, rather than relying on heavy law enforcement. • Guiding design and R&D strategy

Work-Area Based on Image Processing and Pattern Recognition I. Current Progress 1. Artwork based Authentication a) Micro Print / fine line base module b) Line Half-tone document image 2. Printing Technique Identification 3. Ink age determination 4. Paper based Authentication II. Future Work-Plan 1. Enhancement of line-HT Document image 2. Modeling of Ink-Color degradation 3. Implementation by Currency Note verification a) Art-work base b) Printing technique c) Ink base

Bank- Cheques Authentication 1) Color Feature Kurtosis of Image Color Gray level Variation Image Hue 2) Back-Ground Artworks Line Quality Measure Fourier Power Spectrum Binary Correlation

Flow Chart of Bank-Cheques Authentication Classification K-mean NN SVM Poly Test 88. 7% 97. 5% RBF 99. 5% 98. 5%

IHT for Line Half –Tone - Better Visualization - Document examination - For reprint - Automatic document Authentication

Proposed Method - Learning based pattern classification technique - More than one NN involved - Finding lpi information

Result m LPI 3 X 3 5 X 5 7 X 7 PSNR SSIM 4 70 K 24. 88 0. 870 26. 77 0. 88 28. 705 0. 89 8 25 0. 87 26. 87 0. 87 27. 69 0. 87 16 24. 93 0. 75 27. 07 0. 89 29. 67 0. 90

Printing Technique Verification Gray Level Dominant Intensity(f 1) Key Tone (f 5) Hole Count(f 2) Structural Edge Roughness(f 7) Area Differences (f 8) Correlation Coefficient (f 9) Color Contrast (f 4) Avg. Hue (f 3) Avg. Color (f 6)

Result Classification K-mean NN SVM Poly Test 92. 7% 99. 5% RBF 99. 9% 99. 6

Ink-age Determination - Pointing Absolute Document Dating -Yellowish and brownish effect on documents with its age. -Gray and color level features are extracted. -5 -decade samples are considered i. e. 30’s, 40’s, 50’s, 60’s and 70’s. - Through MLP-NN used for determination Dated as 30's 40's 50's 60's 70's Samples 30’s 25 5 6 2 2 40’s 5 27 5 3 50’s 1 4 31 4 60’s 3 32 5 70’s 6 34

Detecting fluorescent paper pulps for Currency Note Checking Overall Approach (i) Detect pulps in a UV scanned banknote under identification and verification phase (ii) Extract features from the detected pulps (iii) Train a NN classifier based training samples that include both genuine and fake notes. (iv) Finally Test the documents by trained classifier.

Enhancement of Low Frequency Line Half-tone Document Block Diagram of Proposed IHT Dot Pattern Corresponding of each quantization level

Result: Enhancement of Low Frequency LHT Document Comparison analysis of Lena Eye a) HT image with different lpi b) reconstructed image c) corresponding PSNR value d) Corresponding SSIM value

Modelling of Ink-Color degradation on old printed documents Role of this model – 1. Better insight about how the color change overtime. 2. Based on ACO based optimization technique 3. Prediction document condition after certain years. 4. Ink age determination Proposed model

Authentication Verify by Paper Currency Note

Result: Authentication of Paper Currency Note Green Line –Printing, Red Line – Ink, Blue Line – Art-Work INK ANN Art-Work SVM Poly Avg. 98. 5% 99. 5% ANN RBF 94. 5% 99. 3% Printing SVM Poly ANN RBF 99. 3% 99% SVM Poly 99. 8% 100 RBF 99. 75%

Publication and Future Plan

Future Plan • Ankush Roy, Biswaiit Halder, Utpal Garain and David Doermann. "Automatic Authentication of Banknotes. " Springer, International journal of document analysis and recognition (IJDAR), (Review stage). • B. Halder, U. Garain, Rajkumar Darbar and Abhoy C. Mondal and “Inverse of Low Resolution Line Halftone Images for Document Inspection”, Springer, 6 th International Workshop on Computational Forensics (IWCF 2014 , ICPR Workshop) Sweden. (Published). • B. Halder, A. C. Mondal and R. Darbar , “Enhancement of low frequency line halftone document images through inverse half toning method”, IJSISE, Inder. Science, (Under Preparation). • B. Halder and A. C. Mondal “Modeling of ink-color degradation on old printed documents ” IJCVR (Under Preparation). • B. Halder, and U. Garain “Microprint line Design based Authentication of Security paper documents”, Elsevier, Pattern Recognition Letters (PRL) (Under Preparation).

Acknowledgements This research work was supported and got technical help from CVPR Dept. , ISI, Kolkata, India. I would like to thank my guides Dr. Utpal Garain (Asso. Prof. CVPR, ISI, Kolkata, India) and Dr. Abhoy Ch. Mondal (Asso. Professor, Dept. Of Computer Science, University Of Burdhwan, WB, India) for his guidance and important suggestions. Also, I am sincerely thank the questioned document examiners of the Central Forensic Science Laboratory (CFSL), Govt. of India for their kind help and cooperation. Thank You