Please enable JavaScript.
Coggle requires JavaScript to display documents.
Single-image DTC - Coggle Diagram
Single-image DTC
Techniques
Re-compression artifacts
Metadata checking
Absence of metadata
Minimal metadata
Specific fields (common, good indicator)
Resolution
Specific dimensions
Editors blacklist
Device (Samsung vs iPhone)
X, 12, 12Pro etc.
S7, S9 etc.
id-document changes
No breaking API change required
features = check_for_absence(reduce_func(metadata))
1: Absence of metadata
2-N: Resolution, specific dimensions
How should IDV treat new features wrt fraud score?
Unit tests
Copy-paste detection (paste for same image)
Stretch tool (pull smile)
Cut-paste detection (paste from external image)
Noise map
Metadata cross-checking
ELA
Geotagging
Gradient checking (edge detection)
Luminance (HSV)
Build, Buy, Replace?
SPIKE (1 day): Services available / APIs / Libraries ?
Cost
Maturity
Functionality (mapping to attack vectors)
Jurisdiction/s
Containers?
Debrief
Technique maturity indicator
Data available (AS IS)
ImageNet (& input features)
JZ to refresh what's been run (AS IS)
Data required
Types
Fake textual data
Fake photo
Fake barcode
Fake insignia etc.
Overlays
Text overlay
Noise overlay
Test data
Copy-paste
Cut-paste (paste from external)
Generated barcode
Text
altered
Look at existing
Resources
links
https://29a.ch/photo-forensics/#forensic-magnifier
https://link.springer.com/chapter/10.1007/978-3-319-97283-1_17
https://blackhat.com/presentations/bh-dc-08/Krawetz/Whitepaper/bh-dc-08-krawetz-WP.pdf
Implementation