A team of European researchers have announced a revolutionary new spam detection technique that promises to change forever how we filter incoming email, and, the researchers claim, could end spam as we know it. I suspect that claim may prove to be overoptimistic, but the technique does sound interesting and their early tests apparently show vanishingly low rates of both missed spams and false positives. What's more, they say that the next version of their detector will reduce the error rate still further.

Briefly summed up, the test hinges on the fact that each piece of email has a unique smell. Using a polymerase chain reaction and an electrostatic precipitation device, the smell of any individual item can be amplified before being fed through a sensitive “electronic nose”. The output signals from the ‘nose’ are then fed into a neural net, which the researchers have trained to distinguish between different types of message.

According to the inventor, French scientist Hervé Poisson d'Avril, the breakthrough came when he applied a prototype device to an email message from his grandmother. “As you might imagine,” he said, “it had a kind of ‘little old lady’ smell — lace and lavender, and just an undertone of Armagnac”. Further tests on messages from other members of his family revealed that each had its own distinctive odor, which could be sampled and recognized using electronic equipment.

The most interesting thing is that while legitimate emails have very distinctive scents that vary from sender to sender, spam messages have a kind of ‘collective stink’ that is both recognizable and unpleasant. So-called 'Nigerian' 419 messages have what Dr Poisson d'Avril calls ‘the smell of Lagos’ — vehicle exhaust fumes and cooking smoke, mixed with decaying garbage. The worst smell, apparently, is that of spams advertising penis enlargement pills, which apparently have “a kind of musty, locker-room smell, like stale sweat and unwashed clothes, coupled with the sharp tang of unrefined sugar”. Another element of the smell is said to be ‘human semen’, to which all I can say is 'Ugh'.

Needless to say, it may be a while before Dr Poisson d'Avril's invention comes to the desktop. Currently, the equipment needed to amplify and then sample the odors takes up most of a small laboratory and is fragile and prone to breaking down. Nevertheless, the researchers believe that in as little as nine months they may have a version suitable for deploying at ISPs, with a 1U rack-mountable version of the device following shortly thereafter. Even if they are never able to miniaturize it sufficiently to fit in an individual desktop computer, let alone a laptop, it's clear that their invention could have a major impact on the way we filter our email.

The team's findings and a description of their invention are published in the April 1st edition of the Communications of the ACM.