There can be lots of forensic evidence on many people’s cell phones when a mass incident like a shooting or bombing is happens, but winnowing out the relevant material and putting it in context can be a time-consuming and tedious affair for law enforcement.
That’s why The University of Alabama in Huntsville (UAH), Florida State University (FSU) and Purdue University have teamed to develop an artificial intelligence (AI) tool to help law enforcement target, extract and collate cell phone evidence related to an incident. The research is funded by a two-year, $600,000 grant from the National Institute of Justice (NIJ).
“So, for example, during the Boston Marathon bombing, several people witnessed the event and had taken videos, etc., on their phones,” says Tathagata Mukherjee, an assistant professor in the Department of Computer Science at UAH, a part of the University of Alabama System. “Law enforcement was given access to this data but had to manually sift through it and create the context for what had happened. Here, we want to use AI to do exactly that to help law enforcement with the investigation.”
The targeted forensics tool under development will not be used in any way for tracking people, Mukherjee emphasizes. Rather, it will be a tool for collecting evidence to be used in court obtained from phones voluntarily provided to law enforcement by witnesses.
“This is pure forensics work. We had developed a targeted phone forensics application on a NIJ award back in 2016, and this work builds upon that and extends the application for multi-phone forensics targeted towards mass incidents like terrorist attacks or mass shootings," said Mukherjee.
That 2016 work is under test by NIJ. It was developed with input from the Florida Department of Law Enforcement, and researchers hope to involve law enforcement in development of the new system, as well.
It’s expected that the new work will also lead to patents. UAH will be responsible for implementing the AI and machine learning algorithms for the endeavor, FSU for the forensics and Purdue for the forensics and systems portion.
“The assumption is that the witnesses to an incident have evidence in their phones and this evidence needs to be extracted and correlated with data from other phones owned by other witnesses,” Mukherjee says.
“Moreover, only evidence related to the incident needs to be extracted, hence it is targeted,” he says. “Naturally, AI needs to be involved in order to implement the targeted aspect of this work and we need to build the whole system for acquiring data from multiple phones and correlating it for creation of the context for the event under consideration.”
UAH’s research will involve a team of graduate computer science students who are yet to be selected.
Republished courtesy of UAH.