Please   note  that  in  the following   website  address:  

http://www.dartmouth.edu/~news/releases/2004/07/01.html

Hany  Farid (Note;   Hany  Farid   is  working   on the  issue  of  
investigating    digital   images)   has  stated    that   original  
photos   are  always   "regular"  but  the  fake   photos   are  
always   "irregular"   so   therefore  I  want  to  mention  that 
those   people  or  companies    which   create   fake  photos   might
 also  be  able  to  create   "REGULARITIES"   in  the   fake   photos
  and  if  that  happens   then   it  means   that  fake   photo(s)  
can  be  "equal"  to  the  original   photo(s).  Please  provide 
"clarification"   to   that  mentioned   statement.

You are almost certainly correct that once Farid or others have
developed an algorithm to deduce whether a photo is original or not,
some fakers will use the same algorithm to try to evade detection; but
Farid's method would nevertheless be useful in detecting those fake
photos where the creators haven't bothered or been able to cover their
tracks. Personally, however, I think the main problem with his
methodology is that very many - perhaps the vast majority - of media
images are digitally enhanced in some way. as compared with their
original state; and I wonder whether his algorithm will be able to
distinguish between such harmless "touching-up" and more sinister
fakery.

Without conferring with the author, it would be difficult to guarantee
that any interpretation accurately describes his findings.

As a personal interpretation based on personal knowledge of digital photography:
Digital images made up of little dots (pixels) of color that are
displayed together with millions of other pixels to create an image. 
Each image contains millions of pixels.  When a photo is edited, the
computer has to "guess" as to how to merge or alter the edited
section.  To the human eye, we might not be able to notice the
difference, but if we examine each individual pixel, we might be able
to identify evidence of computer editing.

For example, if you look at a room that has wallpaper, you probably
will not be able to see the seams between sheets from a distance, but
if spend time up close to the wall, you probably could figure out
where the seams between the pieces are by noticing the gap or slight
shift in the pattern.

The computer is like is able to take a very close look at the lowest
level of detail and look for those gaps or shifts in the natural
pattern to determine that the picture has been altered.  These would
be the "Irregularities" that it can use to detect altering.

As programs are written to look for these gaps, there probably will be
other people working to update image editting programs that alter
images to eliminate the gaps.  Sort of like the professional who
trains to hang wallpaper with less noticable gaps.  It is a cycle
where as the people looking get better at looking for these
irregularities, the people editting the images get better are hiding
their changes and eliminating any of the signs of edditing.

Eventually the people creating fakes will do it so well, computers
might not be able to detect a difference.

Again this is a personal interpritation - please feel free to offer
any alternative views.

In my forays into contract computer graphics, I've done quite a bit of
digital fakery, with no genuine intent to deceive.

I believe it is entirely likely that the day will come when
photographs will no longer have any legal standing as evidence. Even
photographic negatives can be meddled with. Detection of such
tampering will soon, I think, become next to impossible.

I am sceptical... for one: where do you draw the line for calling
something "tampering"? And what is considered a natural image? Is a
digital image of me in front of a house regular/natural? How about me
in front of a poster of a house? How about me professionally "cut out"
of the background and superimposed on the digital image of a house?
The author doesn't give any details on limits/conditions.
Sure I can imagine situations where a "cut out" would show in a
Fourier transform because the edges were too sharp, but that means it
hasn't been done very professionally. I am up to the challenge of
producing a very fake image in 5 minutes that would go undetected...
so, Hani?