Untwirling a Photoshopped Photo

So, this pedophile posts photos of himself with young boys, but obscures his face with the Photoshop "twirl" tool. Turns out that the transformation isn't lossy, and that you can untwirl his face.

He was caught in Thailand.

Moral: Don't blindly trust technology; you need to really know what it's doing.

Posted on October 26, 2007 at 6:44 AM • 43 Comments

Comments

Evil BOctober 26, 2007 7:15 AM

Just a thought; we've established that this is a reversible filter. Could a suitably evil paedophile photoshop someone else's face onto their pictures, then use the swirl filter to to hide it? This has the advantage over plain photoshpping that the damage done by the filter "explains" many of the likely inconsistencies introduced by a simple cut and paste job.

alforaOctober 26, 2007 7:19 AM

The pedophile's problem (except being a pedophile in the first place) was not that he trusted technology but that he _wanted_ to show _himself_ next to the abused children. Just cropping the picture to remove himself or painting a black rectangle over him was out of the question for him.

Obviously he couldn't stand the thought of being "removed" from the picture so instead of choosing easy methods he chose a more sophisticated filter to "stay in the picture". What a sick mind...

Hieronymous CowherdOctober 26, 2007 8:04 AM

There's a point to be made here about reversible and non-reversible algorithms. Choosing twirl to try and erase data is about as sensible as encrypting your /etc/passwd with a Caesar cypher. If he'd used Gaussian Blur he'd probably still be free.

LeoNerdOctober 26, 2007 8:07 AM

I wonder if it would be possible to use some sort of zero knowledge proof, or hashing algorithm, to come up with a way to obscure a picture such that you could prove it was still you, but nobody else could identify who it was.

Trichinosis USAOctober 26, 2007 8:07 AM

People who make a lifestyle of abusing what power they are given have a natural, fundamental and incurable lack of understanding of what that power is really for.

It causes them to make mistakes. Sooner or later, it allows them to be beaten. The trick is not to become a casualty while waiting (with variable amounts of either patience or participation in the acceleration of their karma) for time and circumstance to bring said fools to their inevitable fate.

rickOctober 26, 2007 8:08 AM


So the police think their using a feature openly-built into a product is 'secret'? Talk about sticking your head in the sand...

Green InkOctober 26, 2007 8:09 AM

You know they were sitting on these photos for three years before they figured out how to “undoctor��? them. There is no sophisticated techno-gummery involved: it’s simply the swirl filter in Photoshop. All they had to do was find the centre of the swirl (which is as simple as looking), drag an elliptical marquee round the area of the swirl and use the swirl slider to unswirl it. Why did it take three years to figure that out?
And here’s the really stupid part: why reveal that they’d undoctored the image at all? They should have just released the clear image and said nothing about how they managed to reveal the face. Now other paedophiles who might have used the same technique to obscure their own identities will use different methods. Investigators should NEVER reveal methodologies in these cases.

geomarkOctober 26, 2007 8:42 AM

@Trichninosis, I like that concept of "acceleration of their karma". But not sure how you would actually achieve that. The fruits of karma pretty well run on their own schedule, don't they?

JosephOctober 26, 2007 8:58 AM

"If he'd used Gaussian Blur he'd probably still be free."

There have been several papers written on methods to undo Gaussian Blurring. I also remember my professor talking about the technique in an advanced image processing class. He should have just used cropping or a black rectangle.

MichOctober 26, 2007 9:12 AM

These sickos are too idiotic to find time to learn these technical stuff as they are too busy looking for preys.

Don't think it does any good here in discussing how or what method he could use so that he wouldn't get caught.

They deserve to get caught !

So, just shut up for a while!

Carlo GrazianiOctober 26, 2007 9:14 AM

On Gaussian blur: It seems to me that using an FFT to turn the convolution into a product would make the inversion pretty straightforward, at least for length scales of more than a few (say 10) pixels.

So I guess the point would be that if you're going to use GB, you shouldn't blur with length scales that are long enough that the Gaussian kernel can be resolved by an FFT. If the photo is high-enough resolution that small-kernel blurs don't obscure the required features, use another technique.

FPOctober 26, 2007 9:28 AM

@LeoNerd: "...algorithm such that you could prove it was still you, but nobody else could identify who it was."

Trivial. Just XOR the pixels that you want to obscure with radom numbers, obtained from a PRNG initialized with your secret key.

Colossal SquidOctober 26, 2007 9:32 AM

That would be 'alleged paedophile' unless he's been convicted.
On reading TFA I see he hasn't:
"Canadian pedophile suspect Christopher Neil..."

Innocent until guilty..isn't there something in your Constitution about that Bruce?

Bruce SchneierOctober 26, 2007 9:42 AM

"That would be 'alleged paedophile' unless he's been convicted."

Of course.

But -- according to people who have actually seen the photos -- he's actually is a paedophile. The mapping to the person in the photographs and the Canadian who was arrested is what's alleged. So I was correct in calling him a "paedophile." I was incorrect in saying that "he was caught."

dhasenanOctober 26, 2007 9:46 AM

@Colossal Squid: The person in the pictures is undoubtedly a pedophile. Whether than person is Christopher Neil is questionable. Bruce did not mention Christopher Neil in the summary.

BhimaOctober 26, 2007 9:58 AM

I am reminded of the other criminal, Crooker, that Bruce reported on that sued a bunch of people over the FBI getting past 'drivelock' on his laptop.

I have to admit they sort of hung him in the press, hope they have the right guy.

Carlo GrazianiOctober 26, 2007 9:59 AM

Moreover, "innocent until proven guilty" is a legal standard for imprisoning someone. It is not a standard which places any kind of constraint on private judgment of innocence or guilt. I am perfectly free to refer to this person as a paedophile, based on my private view of what the evidence shows he did. Doing this may or may not create a libel problem, depending on where you live. But claiming that everyone must necessarily honor the legal standard for burden of proof in what they say and write is silly. That's not what that standard is for.

jeffOctober 26, 2007 10:08 AM

@Colossal Squid

>Innocent until guilty..isn't there something in your Constitution about that Bruce?

Actually, there's not. It is, however, a common law principle. However, it only applies to the government. Subject to potential claims for libel and slander (where truth is a defense), Bruce is perfectly free to call that guy a pedophile.

jeff

TrueOctober 26, 2007 10:09 AM

I agree with Carlo. Requiring everyone to honor innocent until proven guilty is to become thought police. Innocent until proven guilty is to protect liberty when there is a reasonable possibility someone didn't do something, it is not to protect one's reputation when they probably did.

AnonymousOctober 26, 2007 10:29 AM

A lot of above people are missing the point. Pedophile is not a charge, it is a psychological disorder. Charges don't stem from being a pedophile, they stem from what you do because of it.

There's no innocent until proven guilty mantra here - on the charges, however, there is. Nobody is found guilty of pedophilia, they're found guilty of sexual interference, statutory rape, child porn, exploitation of a minor etc. - your state/country may vary.

ScottOctober 26, 2007 10:38 AM

Just to be sure... are those links worksafe? I'd hate to find out the hard way that they're not.

Shachar ShemeshOctober 26, 2007 10:41 AM

Joseph mentioned it, but I wanted to clarify.

Even if you use a truly lossy filter (say, "pixelize"), you can rarely be sure that its "lossy enough". It's fairly easy (FFT, Nyquist) to asses whether a single picture contains enough information to be unobfuscated, but the moment we are talking more than one picture (say, video), all bets are, again, off.

Unless you can say the picture contains zero info (say - you blacked out everything that was incriminating), you just can't be sure.

Shachar

John RidleyOctober 26, 2007 11:01 AM

@Scott; yes, the links are safe for work. I'm sure Bruce would have labelled them NSFW if they weren't SFW.

kybernetikosOctober 26, 2007 11:02 AM

>I wonder if it would be possible to use some sort of zero knowledge proof, or hashing algorithm, to come up with a way to obscure a picture such that you could prove it was still you, but nobody else could identify who it was.

Well, clearly you can do it by encrypting the face section of the image with a private key that you have.

But what I think you mean is more or less the same as leaving your face in, unobscured - only people who already know what you look like are able to tell that it's you in the picture. In this case the problem is simply that the government know what you look like.

FNORDOctober 26, 2007 11:54 AM

@Trichinosis USA
You're missing the point. He doesn't make mistakes (only) because he has a sick mind. He makes mistakes because they are inevitable, and one mistake can get him caught. Especially when he makes his crime public.

It's the same problem that software makers have in preventing exploits. No matter how careful you are, eventually a zero-day exploit will sneak through.

tk.October 26, 2007 12:07 PM

"Don't think it does any good here in discussing how or what method he could use so that he wouldn't get caught.

They deserve to get caught !"

While I agree that, in general, people doing bad things should be stopped and whatnot, I think discussing security methodologies is a lot of what this site is about; the applications go further. For example: if you want to post a picture on your blog but you don't want your boss/stalker/ex-girlfriend to know it's you or know your face: don't use the twirl filter.

Personally, I find the eraser to be an effective tool for erasing stuff.

Carlo GrazianiOctober 26, 2007 12:29 PM

I think all would agree that what this guy did is the symbolic equivalent of pasting the face of Mad Magazine's Alfred E. Newman over his own face in those photos.

Ironically, he would have been much safer he'd actually done that.

Trichinosis USAOctober 26, 2007 12:37 PM

@Fnord

Not missing anything. It's a mistake to abuse power. A person who makes that mistake not only will make others, s/he will make THAT one repeatedly. Arrogance is it's own reward.

MathFoxOctober 26, 2007 12:51 PM

If you pixelize your face to 4x4 pixel blocks and I have 16 different pixelized pictures of your face from (more or less) the same direction, would I be able to reconstruct your face to 1x1 pixel resolution?

90_percent_moeOctober 26, 2007 12:54 PM

On twirling: There is no "untwirl" feature. But you can reverse the direction of the spin, and *reapply* the twirl filter.

I just twirled a section of a photo, saved, reopened and applied the same settings, only reversed the spin. Not even close to resembling the org photo. I tried flipping the image horz, in order to reverse the spin that way, no luck there, either. I'll bet it did take them a while to do this--it's hard. Even if you do know the exact settings of the twirl brush (there are four parameters), it's not a matter of simply "untwirling".

EamOctober 26, 2007 1:09 PM

@MathFox:
Interesting thought. It may be possible, but I'd think it would be very unreliable.

I know it's not quite so simple, but we can say you're losing 15/16ths of the information on the face. I imagine that a lot of people's faces would be able to match with that remaining 1/16th.

Paul RenaultOctober 26, 2007 1:10 PM

In Scientific American, at least six years ago, there was an article about the reversibility of entropy.

The author(s) of the article had two clear cylinders of different diameters, one inside the other - the smaller cylinder had a crank/handle on it.

In between the two cylinders, they had different stripes of coloured goop. They rotated the inner cylinder until it looked like the goop was 'unrecoverably' mixed. Reversing the rotation restored the coloured stripes of goop surprisingly well.

I couldn't find the SciAm article but here's a link pointing to a Harvard University description of the experiment.
"Reversible Fluid Mixing"
http://tinyurl.com/34vslz

I guess he

Carlo GrazianiOctober 26, 2007 1:12 PM

Hi MathFox.

I think that's formally correct, if the pictures are identical, and if the pixelization offset is different in each of the 16 photos. The result then would be to reduce the dimension of the null-space of the linear transformation from 15 (per pixel block, per color channel) to zero, in a composite map that would then be invertible (and 16-dimensional block-diagonal, so trivially invertible).

If it really were multiple images from different angles, I can only imagine that it would be a much harder problem. Although, Shachar Shemesh's comment about video (above) would seem to suggest that assumptions of continuity in time across video frames can be used to impose strong constraints that lift the inversion degeneracy.

But I don't do this for a living, so I write under correction.

Jack C LiptonOctober 26, 2007 2:09 PM

A couple of points:

1) The paedophile isn't doing this in a vacuum, he was counting coup.

2) He was, I think, more into "power", and, AFAICT, was mostly turned on by ego.

3) As Londo Mollari said: "Arrogance and stupidity, all in one package... how efficient of you!" I sometimes think that the two are redundant, for instance, given that it's hard to be _truly_ intelligent (and possess empathy) and arrogant... and vice versa. They fuel each other.

4) Children, as near as I can ascertain, are being taught more and more to show obedience to adults. I do NOT think this is healthy and have preferred to live with some level of back-talk from my children as the price of thinking for themselves. I am *not* impressed by blind obedience in children or adults.

5) "It *doesn't* take all kind, we simply *HAVE* all kinds" - stolen .sig that seems to apply.

6) (number 6 has escaped)

7) I am told that paedophilia is not all that scarce, it's the people who ACT on their attraction that are. It's the acting on the desires that crosses the line. It don't matter if you're some random adult or a priest or police officer...

8) If people can be charged with THINKING or DESIRING something illegal, why isn't the Shrub properly institutionalized for his fantasies of Bondage and Discipline, Dominance and Submission, and the like?

9) This crap ain't going to go away because human beings are not "all of one piece", so these kinds of arrogant abusers aren't going to go away... they'll just be elected to office, somehow.

A _lot_ of this publicity will likely get mis-used by our current "leadership" simply to provide something extra to fear... and will promptly mis-use and ab-use any new powers that get wrung out of the legislature.

JosephOctober 26, 2007 4:07 PM

"reversibility of entropy".

That is a fascinating experiment, and it has interesting implications.

But it isn't reversing entropy.

MaltheosOctober 26, 2007 4:19 PM

True entropy cannot be reversed. The methods that we use to aproximate entropy and the methods we use to mix entropy into our ordered structures, those can be reversed ( or at least worked arround)

AnonymousOctober 27, 2007 1:44 PM

I would say "reversibility of apparent entropy".

As I recall the two-cylinders demo, it also wasn't perfectly reversible. But approximately reversible is often good enough. Beyond that, it's just engineering tradeoffs.

guvn'rOctober 29, 2007 8:39 AM

@Fnord, "He makes mistakes because they are inevitable, and one mistake can get him caught."

Exactly. Mistakes are not just karma, they are part of being human.

Years ago a co-worker taught me a great philosophy: "I'm human. I screw up. So I plan for it and try not to let it hurt me."

It's stood me well in many different guises.

Valdis KletnieksOctober 30, 2007 1:06 AM

@Jack C. Lipton:

"3) As Londo Mollari said: "Arrogance and stupidity, all in one package... how efficient of you!" I sometimes think that the two are redundant, for instance, given that it's hard to be _truly_ intelligent (and possess empathy) and arrogant... and vice versa. They fuel each other."

The fatal flaw with this is that you're making the assumption that the person possesses empathy. There's nothing at all difficult about being an intelligent, arrogant, sociopath.

And in fact, I'll wager that there's a lot of people walking around with similar urges who *don't* act on them precisely because they *do* possess empathy for they hypothetical victims - and that the ones who carry through do so because they lack such empathy.

Jack C LiptonOctober 30, 2007 9:43 PM

@Valdis
"The fatal flaw with this is that you're making the assumption that the person possesses empathy. There's nothing at all difficult about being an intelligent, arrogant, sociopath."

Well... intelligence, to my way of thinking, requires the ability to acknowledge that other people EXIST as people-- sociopaths are, to my eye, solipsists, and the rest of us are just cardboard cut-outs. I may be wrong but I want to believe that the truly intelligent have a component of empathy, since, without it, a sociopath has a LOT of extra blind spots and holes in their ability to analyze.

Unfortunately, you cannot teach empathy to a sociopath and the ages-old check-and-balance of duelling is long dead in the US "culture", so there is no simple means of culling the herd. We've got quite a bumper crop these days.

Given Halloween tomorrow, it's easy to forget that the true monsters looks just like any other human being.

Leave a comment

Allowed HTML: <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre>

Photo of Bruce Schneier by Per Ervland.

Schneier on Security is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc..