Originally posted by jeep3r
Thanks for that, Pinke.
Originally posted by ArMaP
Originally posted by jeep3r
P.S.: Would you mind telling us how exactly you went about to achieve that result? Which tools did you use? Thanks in advance ... !
Check your messages.
Why the secrecy Armap? There is quite a few real obvious ways to fool these types of things and not even hard or super secret.
I suppose it is one of my frustrations sometimes ... often persons want to treat these things like secret pixie magic or some such. Like it is back to
the 80s when these things were arcane arts in case the criminals caught on.
Is nothing particularly against you Armap, you're normally very nice and awfully helpful ... but this kind of stuff isn't a big secret! And persons
shouldn't be afraid of sharing those things.
With signal / image processing, ultimately if persons don't understand it, it simply loses credibility as a method. Yes, persons can learn ways to
hide from the techniques but ultimately they do not know if what they've done has left more evidence behind. Is better that persons understand a
methods failings at the very least so they don't create false positives.
To give an idea, it is quite common in VFX to perform operations such as grain / noise matching as a final step, since all elements in the scene will
require it to look similar. It also tends to screw up ELA as a sole method of finding fake imagery and these are persons who aren't even trying.
Original image from OP as thumbnail.
Me being a dork with clone stamp:
Results of ELA with clone stamp:
Results of 'counter forensic' attacks using Median filter
and also additive Gaussian noise. You
can find both these things in photoshop to play with.
Essentially this is making the compression harder to quantify. Even if you're suspicious that the image has been altered, it is harder to tell where
and may even look like a color alteration rather than something sinister on first glance. Median and adding noise together is quite effective and
breaking up block based compressions.
Using two methods that have synergy in counter-forensics is often called a dual-path attack. Median and noise are often used together or alone to
conceal things like resampling and dequantization. There can be a trade off sometimes, since the attack it self can leave behind evidence or clues.
Generally speaking the stronger the attack, the more likely it is to produce another statistical anomaly.
It's why I suppose it's good to use multiple techniques when investigating ... and it makes it especially dangerous to alter high profile photography
like Curiosity since multiple techniques used at the same time can result in the counter technique tripping over itself since new techniques are
always forming. Is why I suggest perhaps that purely CGI based off survey data or something claiming a modified camera sensor or some such
be safer since it will at least be internally consistent.
And blah ... sorry if I sound cranky, but this sort of thing as technique for investigation is a total waste of time if persons hide it under carpets
or are afraid of what persons will say about their techniques. Without helping persons understand why X or Y works its just magic and we may as well
just pull rabbits out of hats whilst screaming 'this rabbit says I am right! Now back off!'
I'm not a very competitive person I guess, and nor am I mega smart. I do like learning tho. In my aweso non-existent fantasy world we all work
together at this stuff, ask questions fearlessly but with humble egos and don't have to watch boyish chest bump sessions to see who is the bestest.
In reality we will be debating lens flares again next week, there will be at least a dozen 'magic image' persons on ATS this year, and I maybe give it
perhaps a few weeks before I get called a name or ten haha
Disclaimer: Pinke is not seriously trying to fake images in this post. Pinke is not intending offensive in this post. Post may or may not be
representative of all Pinke's knowledge. Post should not be consumed. If post contacts with eyes, rinse with cold water. Pinke is an amateur. Pinke
likes leopards and hair dye.
edit on 8-3-2013 by Pinke because: Disclaimers