[Cryptography] Photojournalists & filmmakers want cameras to be encrypted

[Tom Mitchell]

On Sat, Dec 17, 2016 at 5:11 AM, Jerry Leichter <leichter at lrw.com> wrote:

> >
> >> Evidentiary is not a consumer camera. My SWAG is evidentiary is
> >> less than 1% of high quality cameras and perhaps 100% of cell phones.
> >
> > It's journos who are asking for this, and their cameras are definitely
> > part of that 1%.  A reporter who can show that her footage isn't
> > tampered can limit the degree to which people are able to credibly deny
> > or discredit her story.
>

 This case is difficult.  Digital signatures only show that the image and
the signature
are linked.   i.e. ownership not veracity.

> Now think evidence again.  Outside of journalism and investigation
> > contexts, you don't know in advance which cameras are going to be used
> > to record something that becomes evidence. A random college student at a
> > demonstration takes footage of an incident of police brutality using
> > whatever camera they happen to be carrying, and either is, or isn't,
> > able to convince reporters/school officials that the incident wasn't
> > faked, or convince police higher-ups that it can't be credibly denied.
>

The solution to both the journalist and the random college student is
supporting evidence.  A phone call to a recording device like 911 has
discovered edited police videos. Time stamps from doors and other
punctuating
events did not match and demonstrated gaps.   Cycle times of stop lights
further validated the bogus edits.


> I disagree.  This is all typical crypto talk, ignoring the realities.
>
> It's trivial these days to produce a photograph - and possible, though
> right now much harder, to produce a video - that will appear to the
> untrained human eye to be genuine.  Even trained human eyes - of which
> there are few - can be fooled.
>

Photoshop in the hands of a modest talent make still images suspect.
Video in the hands of organizations quickly edit a detail to show the glint
of a weapon justifying deadly force.


> It's also trivial to argue successfully that the photo or video doesn't
> show what it appears to show.  How many videos of cops shooting a suspect
> lying face down on the ground being shot in the head draw the response "You
> can't just look at the picture, you need to understand the totality of the
> circumstances?"  Those responses are made exactly because, most of the
> time, they actually work.  (I'm not here commenting on the *validity* of
> the argument - just its effectiveness.)
>

The issue of story tellers convincing an audience that "you cannot just
look..." can only be countered by
equally skilled story tellers.   AND supporting evidence from multiple
sources.


On the other hand, when it's *really* necessary to check whether a photo or
> video has been tampered with, there are a variety of very sophisticated
> techniques to do so.  The editing technology we have is designed to fool
> the human eye and brain.  It's not designed to fool an analysis that
> examines stuff at the pixel level.  Sure, your eye may be unable to notice
> that part of the background has been duplicated to cover the hole left by
> removing a picture element; but it'll pop right out of an autocorrelation.
> A trained eye *might* notice that the apparent light sources for two
> picture elements aren't at *quite* the same point; but light source
> analysis algorithms will spot this easily and prove that, no, those two
> guys were not actually standing next to each other when the picture was
> taken.
>

Good but uncommon stuff.


> The real threat is *not* the fake pictures - which in important cases can
> be validated.  It's, ironically, exactly what the PBA spokesman brings up:
> The selection of what to photograph and which photographs to present.  The
> individual photographs themselves are completely unmanipulated, but the
> narrative they present is false.  Defending against this kind of thing is
> not a technological problem - though in fact in many situations, technology
> *does* defend against it as a side-effect:  The broad distribution of
> cameras means that any selective presentation is likely to be answered by
> other pictures taken by many other witnesses.
>

Presentation is so important.   One insurance company has a small set of
commercials where the
same lines are delivered in two contexts... One is her joy and the second
is his despair.   A transcript
of the two is identical.  It is too often the best story teller that wins.
   I was listening to NPR and noted
that the audio I was listening to had a lo level background of audio
alarms.  The alarms were not in
the original.   Red breaking  news vs... calmer image of trees on the other
network.


One can come up with situations where a cryptographically verified "seal"
> on a picture - which to be useful must include the time and location -
> would be useful.  Consider photos taken by red light cameras.  Their
> location is fixed and known, and the photos contain a timestamp.  If there
> were reason to believe the photos were being manipulated, an cryptographic
> seal by a third party would be valuable.  (Right now, it seems unlikely
> anyone would bother.  The cheating around these cameras is done in much
> cheaper and easier ways.)  A more realistic scenario is the cameras put in
> place to monitor nuclear installations as part of treaty agreements to
> prevent diversion of bomb material.
>

Oh red light cameras...  that is yet another tangle.
Nuclear installations, spot on ...  worthy of a lot more attention than
just this.


> But I'll contend that such situation are few and far between.  Sure, the
> necessary technology is pretty cheap these days.  If you're talking either
> pro-sumer of professional level camera equipment, or high-end smart phones,
> the additional cost would be minimal.  But I suspect interest would wane
> (if it were ever there) when people realized that they'd taken tens of
> thousands of shots - and no one had ever needed to check the signatures.
>

How does one check the digital signature of an 8"x10" glossy taken with a
20-36MP camera?
The images handed to me sitting in the box a couple years back were B&W
pictures.  When I pointed
to something  that was in conflict with the narrative and passed it to the
next you should have seen the look
on the prosecutor's face and he quickly made a revision that the defense
ignored. ...
Evidence rules today needs revisions for current tech...   As a minimum
multiple cameras better sensors
and more pixels.

The night tactics involve blinding spot lights that even the best 14bit
deep sensors cannot see into the flash or gloom.
A photographer works even paints with light... Today squad cars erase with
light.

This is interesting stuff.

I think it pays to prop up a recording cell phone and step away from it if
you suspect you might witness trouble.
Take cover...
Auto upload to the cloud, auto share with a select short list ...





-- 
  T o m    M i t c h e l l
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20161218/dae9c9e6/attachment.html>