DHS-sponsored research targets social media, gang graffiti

Dozens of federal, state and local law enforcement and first-responder agencies are working with the Department of Homeland Security and researchers from Purdue University to produce new visual analysis technologies for analyzing social media posts, blogs and even gang graffiti. The technologies are now on the verge of adoption nationwide.

Researchers from the Visual Analytics for Command, Control and Interoperability Environments project, known as VACCINE, have developed a tool called ScatterBlogs that allows law enforcement map social media posts in real-time to crime analysis tools, and also enables first responders to better understand what is happening in communities during a disaster.

“We have the ability to take Flickr, YouTube and Twitter feeds and from that create the ability to interactively explore and navigate the topics and anomalies that are occurring,” said Dr. David S. Ebert, director of the VACCINE program at Purdue. But Ebert, who spoke Feb. 11 at the First Responder Technology Demonstration, sponsored by the DHS Science and Technology Directorate, said it’s not about tracking the most popular social media posts.

“If I just show you the top 10, it’s not going to be very useful,” Ebert said. “But if there are 10 tweets that involve a knife, that’s probably going to be unusual, and that’s what a law enforcement officer wants to do.”

The tool, which is currently part of a pilot project with the Ohio State Highway Patrol, is capable of conducting interactive topic classification and interactive anomaly detection of real-time social media messages. It also allows users to filter specific topics of interest, plots the social media postings on a map and allows officials to view profiles and previous posts to help determine a person’s credibility.

Purdue researchers have been refining the tool with direct input from 40 federal, state and local agencies, focusing on everything from crime analysis to monitoring disaster response efforts.

“For example, after Hurricane Sandy, lower Manhattan wasn’t well evacuated after the order went in, but the majority of tweets were coming from grocery stores,” Ebert said. “So people were actually hunkering down to ride it out instead of evacuating. If you were able to pull up that information in real-time, you could decide that maybe you need to get a different message out, or your planning and response might change.”

Researchers are now studying how to integrate other social media outlets into the tool, depending on the feedback they receive from law enforcement agencies.

Fighting gang violence

Another project that researchers from the VACCINE program have developed includes automatic image recognition and interpretation currently being used by more than 15 state and local police departments to document, map and analyze gang graffiti.

Known as GARI, or Gang Graffiti Automatic Recognition and Interpretation, the system allows police officers “to take a picture, categorize and compare the graffiti images that you have to a database, see where there are similar images and do simple shape interpretation from that,” Ebert said.

The GARI system has been pilot tested with the Indiana Division of Homeland Security, the Indiana Metropolitan Police Department, the State of Indiana Gang Intelligence Network and 15 other police departments. More than 8,000 images of gang graffiti have been uploaded into the system since 2012.

Purdue and DHS are currently looking at ways to deploy the technology on a broader basis to U.S. Customs and Border Patrol, intelligence analysts and the nationwide network of fusion centers.

The next steps for the program include developing a community policing version for public access, development of a tattoo image version for use by prison officials, explore use by school districts and find a corporate or government partner to transition the technology to the nationwide market.

Video: How the GARI gang graffiti tool started out as a HAZMAT identification tool