SLIDE 1
Y A N N I C K S C H E E L E N J O P V A N D E R L E L I E
Camera identification on YouTube Y A N N I C K S C H E E L E N J - - PowerPoint PPT Presentation
Camera identification on YouTube Y A N N I C K S C H E E L E N J - - PowerPoint PPT Presentation
Camera identification on YouTube Y A N N I C K S C H E E L E N J O P V A N D E R L E L I E Introduction Why camera identification? Agenda Pattern noise Video encoding Experiment Results Analysis Conclusion Noise
SLIDE 2
SLIDE 3
Agenda
Pattern noise Video encoding Experiment Results Analysis Conclusion
SLIDE 4
Noise sources
Signal processing of a simplified digital camera
Source: FIDIS “D6.8b: Identification of images”
SLIDE 5
Pattern noise
Present on all frames Fixed pattern noise (FPN)
Defective pixels
Photo Response Non-Uniformity (PRNU)
SLIDE 6
Algorithm
SLIDE 7
Algorithm
Correlation between the reference pattern and the
video pattern
Correlation on each color channel (RGB)
Sum of correlation on each color channel
Correlation value between -3 and 3
SLIDE 8
PRNUCompare
Algorithm implemented in PRNUCompare
Developed by NFI (Netherlands Forensics Institute)
http://prnucompare.sourceforge.net/
SLIDE 9
PRNUCompare
SLIDE 10
Video encoding
Advanced Video Codec (AVC) Compresses the video stream Modifies the pattern noise Applies to YouTube
SLIDE 11
Research question
How does re-encoding the video with the Advanced Video Codec influence the pattern noise?
SLIDE 12
Experiment
5 different camera models
Canon Ixus/SX210 Panasonic FP7/FZ45 Apple iPhone 4
5 different cameras per model Multiple resolutions
640x480 1280x720
SLIDE 13
Experiment
1 reference video per camera per resolution 1 natural video per camera per resolution re-encode each natural video
AVC encoding setting: CRF 18,21,…,39
Upload/download videos to/from YouTube
SLIDE 14
Encoding quality 18
SLIDE 15
Encoding quality 21
SLIDE 16
Encoding quality 24
SLIDE 17
Encoding quality 27
SLIDE 18
Encoding quality 30
SLIDE 19
Encoding quality 33
SLIDE 20
Encoding quality 36
SLIDE 21
Encoding quality 39
SLIDE 22
Results
Extracting the pattern noise for each video Correlate each video to the reference patterns Total number of videos processed: 835
SLIDE 23
Analysis
Verify that pattern noise can be used for source
identification before re-encoding
SLIDE 24
Analysis
SLIDE 25
Analysis
Correlation between re-encoded videos and
reference patterns
SLIDE 26
Analysis
SLIDE 27
Analysis
SLIDE 28
Conclusion
Depends on the level of compression Presence of pattern noise differs per model Higher resolutions videos perform better
More pixels == more noise
SLIDE 29
Conclusion
Even after a re-encode on the video with a compression similar to YouTube, it is still possible to identify the source camera for most cameras.
SLIDE 30