More on Smell Samples

Earlier this month, I blogged about a library of people's smells kept by the former East German police. Seems that the current German police is still doing it:

The Stasi secret police used scent gathering in Communist East Germany, collecting smells in empty jam jars and storing them. The method has reminded Germans of that failed regime of snoopers, and was highlighted in the recent Oscar-winning film "The Lives of Others" about a Stasi surveillance officer.

The domestic policy spokesman for the Social Democrat Party, Dieter Wiefelspütz, finds the new weapon "pretty bizarre." But he knows that unappetising though it may be, the method has been employed by German investigators for a long time.

In legal terms, recording someone's body odour is no different than taking their finger prints. It's covered by the criminal statue book. The scent contains a person's identity just like the lines of his finger tips or his DNA.

Taking someone's DNA is subject to strict conditions but the law permits finger printing and scent recording whenever police deem it necessary as part of a criminal investigation -- which means virtually always. Erhard Denninger, an expert on Germany's justice system, has no problem with scent analysis. "It's harmless by comparison with sledgehammer plans like searching people's computers," he said.

Suspects are told to hold several 10 centimeter steel pipes in succession for several minutes each.

There are strict rules governing this procedure. The interior minister of the state of North Rhine-Westphalia has decreed that "persons must contaminate the metal tubes through their hands", and that the aromatic traces thereby recorded "be secured in glass containers in dry condition."

It sounds harmless. But a number of defence lawyers, Düsseldorf-based Udo Vetter among them, advise their clients not to agree to scent recording. If the state sniffs the sweat of its citizens, it amounts to a "considerable intrusion into one's intimate sphere," he says.

The complexity of collecting someone's scent is the theme of Patrick Süskind's novel "Perfume", recently made into a movie, in which an 18th century murderer wraps beautiful women in cloths which he later boils. Unlike in real life, the perfume specialist chose to kill his victims before taking their scent.

Posted on August 1, 2007 at 2:05 PM • 12 Comments

Comments

HenryAugust 1, 2007 3:23 PM

Well, the method used now by (west) German police looks kind of similar on first glance to the old one by the East-German Stasi (Ministery of State Security - Ministerium für Staatssicherheit), but in fact the differences are striking.

East-German Stasi build up a permanent library of samples from known dissidents, that could be used in the future to identify objects in their possession of maybe train dogs to search and identify them.

In the new case police took some samples from a small number of identified suspects in one defined criminal case, where they had an object that was touched but did not carry fingerprints or usable DNA-samples. Those samples were used only for comparison with the deposit investigating this one case and destroyed immediately afterwards.

BetaAugust 1, 2007 3:26 PM

"Nevertheless, German investigators rely on a strictly defined scent identification procedure, one that has been put to the test thousands of times. Six tubes are placed on a podium, including the sample belonging to the suspect. The dog has to sniff an object that has been handled by someone other than the suspect and then has to find their sample.

"In the final test, six samples are laid out, but not the one from the first test. Now the dog has to sniff the real evidence, maybe a tool used for a break in. If three dogs come up with the same results in the preliminary and main tests, then the failure rate is one in every 1.2 million. That at least is what a researcher at the University of Paderborn has calculated.

WHAT? Either that nameless "researcher" has access to a lot of evidence not mentioned here, or else this test is worthless. Since there's no mention of control tests (when the person who handled the object isn't among the six), it simply indicates which of the six people smells most like the criminal. And even if we assume that the culprit must be among the six, if the dogs chose randomly they'd damn the suspect once in 46656 times, not 1.2 million.

RoyAugust 1, 2007 3:41 PM

This has to operate on the basis of gullibility. If you are foolish enough to believe that your scent is permanent throughout life, and that once captured, it can identify you for the rest of your life, then you will fear touching anything that could ever get you into trouble.

Try eating excessive amounts of garlic, cilantro, or tomatos for a few days running and even your nose can detect the changes in your skin scent.

Germany and the US are two of the few countries in the world where the bulk of people believe dogs can accurately identify on the basis of scent alone. The rest of the world knows were stark raving mad.

MRAugust 1, 2007 3:49 PM

This begs the question, when will smell become a viable authentication mechanism for computer systems? ;)

Jamie FlournoyAugust 1, 2007 3:58 PM

>The rest of the world knows were stark raving mad.

Except the Japanese, who are working on a robot dog for scent identification purposes.

No, not really, but it's believable isn't it?

GrahameAugust 1, 2007 6:16 PM

ahhhh, is there any published work that subjects this kind of thing to bayesian analysis?

HansAugust 2, 2007 5:44 AM

Roy: Probably you are the only one here foolish enough to believe that your scent is not permanent. Let me share a fact to you: The part of it that dogs are trained to identified is very permanent and will not be affected by what you eat or do. Sure, your scent might smell different to you depending on what you eat, but what identifies you is still there and dogs are trained to recognize it. The east-german Stasi had extensive research facilities and they have tested such things and found smell samples to be working very reliable.

So go around and touch what you like if you prefer to believe you are secure while you aren't.

BetaAugust 2, 2007 8:31 AM

#Jamie Flournoy: "...the Japanese... are working on a robot dog for scent identification purposes.

"No, not really, but it's believable isn't it?"

More than believable, it's inevitable. Sooner or later someone (not neccessarily the Japanese) will build a machine that can quantify these scents. Then we'll have databases, sweeps, sensors in train stations, the whole nine yards. And the science and scepticism may not be any better than what we've seen for polygraph machines.

Matthew SkalaAugust 2, 2007 10:38 AM

"when will smell become a viable authentication mechanism for computer systems?"

Used in my NaNoWriMo novel a few years ago. I even got in the obvious pun on what you'd call defeating a smell-based lock.

http://ansuz.sooke.bc.ca/books/fff/fff-fulltext.php

Turn cookies off before visiting that page, or it won't let you in. Does this count as "publication" and eliminate the possibility of anyone patenting it?

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 Resilient Systems, Inc.