This report has explored the possibilities and constraints of using physical micro-structures for identification. Utilizing the physical structure itself from an object negates the requirement to add a special marker for identification. Buchanan et. al. [10] discovered that paper documents, packaging and plastic cards contain microscopic surface structures which are unique for the sample. The naturally occurring randomness forms both a unique and currently unclonable identification token. The project deals with 2 major research questions. Part one of this report is dedicated to an imaging algorithm that always extracts exactly the same patch of micro-structure from an acquired sample, independent of the way the acquisition was done. The next part of this project report relates to the identification system as a whole and examines the main bounds of an identification system depending on micro-structures….
Contents: Framework for robust forensic image identification
1 Introduction
1.1 Identification and Forensics
1.2 Scenario
1.2.1 System Architecture
1.2.2 Acquisition and Synchronization
1.2.3 Micro-structures
1.2.4 Digital Fingerprinting
1.2.5 Identification
1.3 Research
1.3.1 Theoretical Framework
1.3.2 Research scope and questions
1.3.3 Research validation
1.4 Thesis Overview
1.5 Contribution
1.6 Notation
I Imaging
2 Image Synchronisation
2.1 Introduction
2.2 Acquisition
2.3 Feature based Registration CONTENTS
2.3.1 On Scale-space and features
2.3.2 Gaussian scale-space
2.3.3 Scale-space derivatives
2.3.4 The Second moment and Hessian matrix
2.3.5 Features points
2.4 Feature based Registration Algorithm
2.4.1 Edge detection
2.4.2 SIFT feature detection
2.4.3 Inferring the projective transformation
2.5 Fourier based Registration
2.6 Post processing and Comparison Metrics
2.6.1 Post processing
2.6.2 Comparison Metrics
2.7 Validation
2.7.1 Feature based synchronisation
2.7.2 Closing Thoughts
2.8 Future Work
II Identi cation
3 Fundamental Aspects of Noisy Databases
3.1 Introduction to Information Theory
3.2 Concepts and Building blocks
3.2.1 Entropy and Mutual Information
3.2.2 The Asymptotic Equipartition Property
3.2.3 The Noisy-Channel Coding Theorem
3.2.4 Jointly Typical Sequences
3.2.5 Random and Typical Set Decoding
3.3 Metadata
3.3.1 Channel Identification Limitations
3.3.2 Concepts and Limitations of Meta-data CONTENTS vii
4 Empirical Identification Limits
4.1 Validation method
4.1.1 Scenarios
4.1.2 Device channel distortion
4.1.3 Entropy
4.1.4 Mutual Information
4.1.5 Intra en inter class distance distribution
4.2 Results
4.2.1 Identical enrollment and identi cation device
4.2.2 Closing thoughts
4.2.3 Device mismatch
5 Future Explorations in Identification
5.1 Introduction
5.2 Cross correlation and Coeficients
5.3 Random Projections
5.3.1 Dimension reduction
5.3.2 Smoothing of the Projector
5.4 Reliable Components and Fast Searching
5.4.1 Random Projections and Magnitude Sorting
5.4.2 Local Variances
5.5 Circular micro-structure extraction
5.5.1 Results
6 Conclusion
6.1 Image Synchronisation
6.2 Identi cation Limits
6.3 Future Work
A Image Features…
Source: University of Twente
Download URL 2: Visit Now