Original URL: http://www.theregister.co.uk/2010/01/20/ibm_security_patents/

Big Brother Blue seeks biometric anti-terror patents

Tracks eyes, breath, aftershave

By Rik Myslewski

Posted in Security, 20th January 2010 06:02 GMT

IBM has filed applications for a dozen patents that seek a whole new level of airport security.

Forget full-body scans, which are currently being touted as the Next Big Thing™ in security in the wake of the Christmas Day Nigerian crotchbomber. We can't even be sure they would have stopped him.

The IBM patent applications, turned up by InformationWeek's Alexander Wolfe, outline technologies that track you as you stroll through an airport, sniff you, check out what you're eating and drinking, scrutinize your attire, watch for "furtive" eye movements, and employ a host of other sensors.

That information is then fed to an inference engine that compares the sensors' data to a knowledge base and alerts security if its set of rules determine you might be a threat. And all of this will be done in real time - or, at least, real enough to nab you before you get on an airliner and activate your lethal underpants.

The core idea of the patents is to use a variety of sensors arrayed throughout an airport, both indoors in out. These sensors attempt to both spot suspicious activity and find already-identified suspects, matched to a knowledge base of possible terrorists.

IW details three of the patents. One, "Detecting Behavioral Deviations by Measuring Eye Movements," describes methods of tracking a subject's eye movements by a battery of cameras looking for "furtive glances, fixed and unblinking stares, concentrated focus, lack of focus, changes in pupil dilation, or any other unusual eye movements." If a subject's eye movements are considered sufficiently shifty, he or she would be identified "as a person of interest" and the gendarmes would move in.

Although the majority of the patent filing deals with identifying as-yet-unknown shifty individuals, the system's cameras could be supplemented with audio recorders and match images and sounds against a knowledge base of known suspects. The system could track a face, license plate, an identification badge, tattoos, scars, an eye's iris pattern, or its retinal pattern. It may also use audio data to identify a person's voice.

A second patent, "Unique Cohort Discovery from Multimodal Sensory Devices," takes the ocular-recognition technology to the next level by expanding the types of sensors to be used to infrared, chemical, biometric, and temperature sensors; motion, metal, and radar detectors; GPS receivers; microphones; photosensors; seismographs; and anemometers. This filing isn't limited to tracking folks in airports, but in all public spaces, including parks and stores.

Automatic 'cohort' detection

The filing is based on the idea of identifying members of predetermined "cohorts" - which it defines as "people or objects that share common characteristics or experience." Some objection may be raised that this amounts to profiling. But since the profiling is performed by machines and not humans, the civil-libertarian objections to profiling could be considered to be minimized.

The individual aspects of a person that might be observed are vast. This includes such things as a subject's age, the make and/or model of their vehicle, the color of their hat, breed of their dog, sound of their vehicle's engine, a medical diagnosis, item of clothing, walking, talking, running, what they're eating, or what they buy.

Event-based cohort identifiers include eating, smoking, walking, jogging, walking a dog, carrying bags, carrying a baby, riding a bicycle, how fast a subject is walking - even the "brand of soda purchased by the cohort." Chemical sensors could be used to identify subjects based on such distinguishing characteristics as their perfume, aftershave, scented shampoo, scented lotion, and other smells. Perhaps one cohort might be identified for its infrequent bathing.

Like the previous filing, this one also has elements that could match subjects with a knowledge base of predetermined suspects, such as license plate, face, and voice recognition.

The third patent filing discussed by IW, "Detecting Behavioral Deviations by Measuring Respiratory Patterns in Cohort Groups," has much in common with the first patent - the shifty-eye detector. Unlike the second, it focuses specifically on transportation systems, namely airplanes, trains, and buses, as places where security personnel would want to collar those troublesome "persons of interest."

This patent filing describes video and audio methods of ascertaining respiratory changes by analyzing changes in movement of the rib cage and abdomen to determine such breathing characteristics as regular, deep, shallow, slow, fast, erratic, or abdominal breathing, or breathing from the chest. If irregularities are found, a "risk assessment engine" will judge the level of risk that it associates with those irregularities.

The key to these three IBM patent filings becoming effective methods of apprehending suicidal evildoers, of course, is not only the sensitivity and accuracy of the sensors, but also the quality of the information in the system's knowledge base. And that's the system's Achilles heel.

First, the cohort definitions need to be realistically set and continually monitored. This might be devilishly difficult in practice, especially when needing to be adjusted not only for such environmental factors as weather and time of day, but also for cultural dissimilarities such as the downcast eyes and nervousness of people simply unaccustomed to either flying or the hustle and bustle of urban airports.

In addition, although all three patent filings focus mostly on identifying subjects who stand out in a crowd due to behavioral patterns, there's also the question of the ability of intelligence services to create a thorough, shared, and reasonably up-to-date database of known suspects.

Intelligence agencies are reasonably good at gathering intelligence, but sharing that information in a unified, useful knowledge base has proven more problematic. As Newsweek reported, for example, after the father of Christmas Day's Umar Farouk Abdulmutallab had notified the American embassy in Abuja, Nigeria, of his son's disappearance, the CIA entered him into the National Counterterrorism Center's Terrorist Identities Datamart Environment database - but never told the FBI about him.

In addition, the UK denied Abdulmutallab a visa, but the US authorities never got that information. They never asked. Also, according to Newsweek, the NCTC and CIA's failure to tell the FBI about Abdulmutallab kept him off that agency's Terrorist Screening Center 400,000-name watch list, or the list of 13,000 people who are put through more-thorough airport screening, or the 4,000-name "no fly" list.

The luckless Nigerian crotchbomber is only one example, to be sure, but his slipping through the system is symptomatic of the isolated silos in which security information is stored. Without a centralized knowledge base, no amount of eyeball scanning nor respiratory-pattern monitoring will stop the next determined wannabe aerial assassin. ®