cnn-video-log-recovery.txt

Extract and patch in additional unredacted data
to cover missing parts of redacted log

The logfile shown in the CNN video is called:
'35200217 su logs 0703141641 to 080310119.csv'

1. Get the 1280x720_3500kB/s videos:
wget http://cnn-f.akamaihd.net/cnn/big/world/2014/05/27/pkg-quest-inmarsat-data.cnn_27020448_,512x288_55,640x360_90,768x432_130,896x504_185,1280x720_350,0k.mp4.csmil/bitrate=4 -o pkg-quest-inmarsat-data.cnn_27020448_1280x720_3500k.mp4
wget http://cnn-f.akamaihd.net/cnn/big/world/2014/05/27/intv-quest-inmarsat-data-dickinson.cnn_27014108_,512x288_55,640x360_90,768x432_130,896x504_185,1280x720_350,0k.mp4.csmil/bitrate=4 -o intv-quest-inmarsat-data-dickinson.cnn_27014108_1280x720_3500k.mp4

(nb. first video had been posted three days earlier at:

  http://ht.cdn.turner.com/cnn/big/international/2014/05/24/quest-inmarsat-data.cnn_24034342_1280x720_3500k.mp4

but subsequently removed).

2. Checksum
sha1sum < *.mp4
d27c18c16058dda4d7b733a6564355e96d861575
966e3d8ae2233cab1677fcb00c588855fd1b1c2c
Length 105290297 (101MB)
Length 76773900 (74MB)

3. Extract the frames
00:51--00:56
03:16--03:20
03:52--03:58
and:
00:41--00:46

VLC option:
$ vlc *.mp4 
Pause
Frame-step using 'e'
Video->Snapshot menu

FFMPEG option:
ffmpeg -i [input.mp4] -ss 00:00:51 -t 00:00:05 -r 1 img_%4d.png

4. Determine Perspective grid (unbend paper)
Not done - Put to one side after a couple of hours experimentation

5. Invert Perspective Map
Not done - used an affine (rotate+skew) transformation in Inkscape

6. Select focus mask based on separate frames
Not done, would require successful 'flattening' first

8. OCR
Eyeballed for the moment, instead of OCR

9. Iterate based on constraints
To do: Decoding the payload to inorder to refine the hex candidates