Thanks for posting the original image. I used the Digital Invisible Ink Toolkit, located at
http://diit.sourceforge.net/index.html. When comparing both smaller images, I found them to be identical. I have a little too much time on my hands... so I decided to keep experimenting...
I wanted to see what kind of difference encrypting "AIM HIGH" in the picture would make.
When I did this I had these results:
Results of benchmark tests
==========================
Average Absolute Difference: 4.6866898009652585E-4
Mean Squared Error: 6.082725060827251E-4
LpNorm: 3.0413625304136254E-4
Laplacian Mean Squared Error: 4.206612200901864E-6
Signal to Noise Ratio: 1.3982824224590164E8
Peak Signal to Noise Ratio: 5.85936396E8
Normalised Cross-Correlation: 0.9999989969072935
Correlation Quality: 126.60633961495463
This shows that the software is capable of finding very small changes in the image... and I decided to make one more attempt. I took the new high quality image, and resized it to match the one they released earlier, as the program requires the images to be the same size. Upon comparison, I had there results:
Results of benchmark tests
==========================
Average Absolute Difference: 15.601511706752822
Mean Squared Error: 756.6327031231303
LpNorm: 378.31635156156517
Laplacian Mean Squared Error: 2.111745474975075
Signal to Noise Ratio: 112.4107839655667
Peak Signal to Noise Ratio: 471.0462533919842
Normalised Cross-Correlation: 0.9982814866037124
Correlation Quality: 126.38889170404684
If you look at the Average Absolute Difference, this, along with most of the other values, is ORDERS OF MAGNITUDE greater than my "AIM HIGH" test. Now, I can't guarantee that my method of resizing the large image would match the method the GDC did, which would cause these differences. Also, if they chose to make 'minor adjustments' to the actual content of the image, that would create this kind of difference in the photo. I think that differences seen here are very high for a simple text message - it would have to be over 1000 characters long.
Regardless, I can tell you that small image and the large image are subtley different. Whether there was some touch-up going on, or if there is a message hidden in the background, I am unable to tell, but they are clearly different at some level.