In the 2002 Tom Cruise movie “Minority Report,” crime in the year 2054 is eliminated by an elite PreCrime squad whose special powers enable them to see the future and predict crime.
It may be 42 years until 2054, but some police agencies have already begun to predict future crime by mining “Big Data.” Big Data is any data that can be accessed legally and potentially assist police in next-generation crime fighting, from phone records, financial transactions, and crime statistics to closed circuit videos, e-mails, and still images.
The idea is not that new. Law enforcement has always collected data. What’s new is the amount of data. Google CEO Eric Schmidt says five exabytes of data is generated every 48 hours. An exabyte is one million terabytes. There’s a lot of crime to fight. According to the FBI, in 2009 there were 1.3 million violent crimes, 9.3 million property crimes, and 6.3 million larceny/thefts. That’s one crime every 2 seconds.
What’s also new is how the data is mined.
“We’re inventing new ways to proactively capture data that may be useful in investigations, and to find smarter ways to cross-reference government data across multiple agencies and data mined in the private sector,” says Philip Becnel, managing partner of Dinolt Becnel & Wells Investigative Group (Washington, D.C.).
Law enforcement agencies in the Washington, D.C., area now collect data from dozens of stationary cameras that capture 1,800 license plate images a minute, information stored for later investigative use. The FBI recently announced it was making its facial recognition technology available to other law enforcement agencies.
“While the FBI has pledged not to mine photographs from social media to add to its current database of around 13 million mug shots, it can and does routinely search the Internet for matches to crime suspects,” Becnel says.
Social media is a mother lode of information. Researchers at the University of Wisconsin recently taught a computer how to detect bullying-related tweets among the 250 million publicly visible messages posted daily on Twitter. Now the machine is identifying more than 15,000 bullying-related tweets most days. This specific tweet traffic seems to ebb on weekends, presumably because school is in session only on weekdays.
The FBI is also developing a leap-ahead technology called Sentinel, a case management system that allows investigators to search for unique pieces of data, like crime details or phone numbers, across multiple investigations.
“Big Data has great potential to predict crime, crime hot spots, and criminal trends,” says John DeCarlo, a professor of criminal justice and forensic science at the University of New Haven. “When we look at technologies like criminal path mapping and predictive policing, we begin to see the same potential in using Big Data that retailers have observed.”
In 2011, the Santa Cruz, Calif., Police Department began using crime records from the city’s archives that contained decades of data on when, where, and what types of crime occurred in the past to predict when, where, and what future crimes may occur.
The Santa Cruz crime model originated with UCLA mathematician George Mohler, who noticed that, over time, crime maps tend to resemble maps of natural phenomenon, like earthquakes. Earthquakes tend to swarm. So do crimes. The trick is predicting when, where, and how many. Mohler adjusted the earthquake algorithms previously used to predict earthquake aftershocks to produce hot spot maps where future burglaries or car thefts might be likely to occur. When Santa Cruz police used Mohler’s maps to deploy patrols to predicted crime hot spots, there was a 27-percent drop in burglaries compared to the same area in the same month the year before.
In August, the New York Police Department, in partnership with Microsoft, launched a state-of-the-art crime prevention and counter-terrorism technology based on Big Data mining, called the Domain Awareness System.
DAS is capable of rapidly blending and analyzing realtime data gathered from roughly 3,000 civic closed-circuit cameras, 911 call recordings, license plate readers, dozens of radiation sensors distributed around the city, as well as historical crime reports. Now the NYPD can now do things like track a vehicle and instantly determine nearly everywhere it’s been for the past few days or weeks; instantly access a suspect’s arrest record, and all the 911 calls related to a particular crime; map criminal history to geospatially and chronologically reveal crime patterns; and verify whether a radiation alarm was triggered by components of a dirty bomb or a harmless medical isotope.
Big blue & Big Data
Last year, IBM raised the Big Data-crime prevention bar by acquiring i2, a UK-based security firm with more than 4,500 customers in scores of countries. Law enforcement agencies and corporate security departments use i2 software to identify fraudulent or improper activity within their logs of operational data.
The i2 Intelligence Analysis package provides law enforcement, defense, government agencies and private sector corporations the ability to harness new intelligence instantly so officials can identify, investigate, predict, prevent and disrupt criminal, terrorist, and fraudulent activities.
Until recently, Big Data was mined mostly for commercial marketing purposes. Amazon, for instance, pushes realtime purchasing suggestions to customers based on music, books or other merchandise that customer has recently browsed or purchased. Data mining will soon evolve to location-based marketing, which targets consumers based on their present location, notifying them on mobile devices of deals in their immediate area.
University researchers in Israel think Big Data mining can be used to predict the location of criminals and terrorists. Like everyone else, criminals (including terrorists) leave digital traces with much of what they do, whether sending or receiving e-mails, making and receiving calls on cell phones, or from credit card transactions. This data can be mined. Engineers at Tel Aviv University are using digital traces to catch criminals and bolster homeland security against the threat of terrorism.
A team of Israeli-based industrial engineers is developing high-powered context-based search algorithms to analyze digital data on-the-fly. The researchers say the new algorithms give investigators the ability to process new pieces of information instantly.
The data used in this work comes from the mobile device and telecom domain. “Even in the age of Big Data, it is extremely hard to discover patterns within the data that bubble up suspicious behaviors,” says Aviv Gruber, lead author of a recent paper describing the Israeli work. Still, patterns found in mobile usage may imply suspicious behavior.
The algorithm works like a computerized Sherlock Holmes. It takes distinct pieces of information like phone calls, e-mails, or credit purchases and reduces them to a set of random variables. The researchers say all of these communications are actually pieces of one long message waiting to be decoded. In a single telephone call, for example, there are several variables to consider—the call recipient, length and caller location. Once all this is known, the algorithm assesses patterns of crime to predict future movements and creates a probability map displaying the possible locations of the person or group of interest.
“The outcome is used in realtime to track the likely location of criminals on-the-move,” Gruber says.
Probable location information may be advantageous in cases where authorities may have just one chance at apprehension. The location of highest probability is the best bet. Zones of lower probability can be ruled out and attention can be focused in increasingly specific areas.
Connecting future dots
Gruber says in the future the Israeli scheme could be used for investigative activities such as pedophile traps. The Internet has given pedophiles a new arena to pursue deviant behavior. Gruber says the Israeli algorithm can be used to automatically track the digital patterns pedophiles leave.
Big Data leads down other roads. Becnel believes soon we’ll develop new ways to connect the dots with the information we already have.
“Imagine the ability to instantly take a security camera photograph from a bank robbery and match it using a facial recognition algorithm to a photograph in an out-of-state motor vehicle database, and then to link that person’s name to a mobile phone from a private-sector marketing database,” Becnel says. An affidavit for a warrant could then be automatically generated, electronically signed and forwarded to a judge. Once granted, investigators could use a code to enter an encrypted portal to the telephone service provider and get GPS coordinates that lead right to the suspect.
“This is all possible today,” Becnel says.
Douglas Page writes about science, technology and medicine from Pine Mountain, Calif. Reach him via firstname.lastname@example.org.