funsec mailing list archives

Re: [privacy] Digital Camera Fingerprints


From: "Marius Gheorghescu" <mariusg () microsoft com>
Date: Mon, 24 Apr 2006 11:48:09 -0700


Although the source is boingboing, that "research" is in fact valid. 

The CMOS sensor noise at low ISO is entirely random but as you increase
the ISO, certain pixels get 'stuck' to maximum response and it's always
the same pixels, assuming the temperature distribution on the sensor
remains constant. They are called hot pixels. P&S cameras have small
sensors and thus they begin to show this template noise much sooner than
professional cameras (at ISO 200 let's say, compared to ISO 1600).  But,
temperature is a big factor here and can change the noise pattern, even
more than the silicon impurities.
A second type of camera fingerprints are the dead pixels - these remain
stuck to 'on' independently of the ISO and exposure length. They are
very rare in general.
A third type of fingerprints is the sensor noise. Dust particles that
attach to the sensor will produce a dark spot on the final image (and
this changes very little depending on the lens settings). For cameras
with inter-changeable lenses dust is a very common problem (but it is
also very easy to remove, thus not much of a fingerprint).

Then there is the type of information that is stored by the camera (time
- which can be correlated), serial # (which is unique), owner
information, firmware version, thumbnail data. A couple years ago there
was an example where one could find edited images by comparing the image
with its embedded thumbnail (certain applications "forgot" to re-save
the thumbnail after the image was edited). They did a scan on the net
and came up with some pretty interesting results. 

In conclusion, there are definitely sensor and camera fingerprints that
can be easily identified given enough images. I wouldn't be surprised if
sensor fingerprint would end up being accepted as evidence in court one
day. 

There is a tool that finds hot/dead pixels but I can't remember its name
right now. Its purpose is to find defects in the sensor of course, not
to extract the fingerprint ;-)

-----Original Message-----
From: Justin Polazzo [mailto:jpolazzo () thesportsauthority com]
Sent: Monday, April 24, 2006 9:25 AM
To: Fergie
Cc: privacy () whitestar linuxbox org
Subject: Re: [privacy] Digital Camera Fingerprints



-----Original Message-----
From: Fergie [mailto:fergdawg () netzero net]
Sent: Monday, April 24, 2006 10:08 AM
To: Justin Polazzo
Cc: privacy () whitestar linuxbox org
Subject: Re: [privacy] Digital Camera Fingerprints

Be on the lookout for tools to "de-fingerprint" digital images --
software which can remove unique pixalization characteristics of
images
related to a given digital camera without significantly degradinging
the
image quality. :-)

In fact, one would think that any given steganography software would
already suffice:

 http://www.jjtc.com/Steganography/

---------------------------------------

And where is the re-de-fingerprinter? I bet some of that altered data
will be stored in metadata by the lesser applications ;-)

-JP
_______________________________________________
privacy mailing list
privacy () whitestar linuxbox org
http://www.whitestar.linuxbox.org/mailman/listinfo/privacy
_______________________________________________
privacy mailing list
privacy () whitestar linuxbox org
http://www.whitestar.linuxbox.org/mailman/listinfo/privacy


Current thread: