SIU prof gets prestigious National Science Foundation grant to enhance AI crime analysis

A researcher at Southern Illinois University Carbondale has received a grant to improve AI crime analytics capabilities while fostering collaboration among multiple agencies and keeping sensitive information secure.

Ahmed Imteaj, assistant professor in the School of Computing, received a $167,500 grant from the National Science Foundation’s Computer and Information Science and Engineering Initiation Initiative. The program, known as CRII, is aimed at providing essential resources for early-career researchers.

Imteaj’s work focuses on creating a federated learning framework specifically tailored for AI-driven crime analytics across interdependent networks. Federated learning – also called collaborative learning – is a subcategory of AI that focuses on creating settings in which multiple clients work together to train a model without transferring sensitive data.

Federated systems have the advantages of securing data privacy, improving efficiency and boosting teamwork across law enforcement agencies, Imteaj said.

“This approach allows agencies to share model updates rather than raw data, and that ensures timely and accurate insights that are crucial for prompt decision-making by law enforcement,” he said. “Federated systems also promote cooperative efforts among different jurisdictions, which makes AI models more robust and leverages the many different datasets for more effective crime prevention and response strategies.”

Imteaj’s framework will include three phases. It will start with data labeling and rebalancing the models to prevent bias, focusing on creating algorithms that ensure fair outcomes across various systems of interconnected nodes.

“Bias in crime analytics can lead to unfair outcomes by perpetuating historical prejudices and exacerbating existing disparities,” Imteaj said. “For example, biased data may result in algorithms unfairly targeting certain neighborhoods, leading to excessive surveillance and enforcement in already marginalized communities.”

To address this, the AI models being developed will incorporate equitable data from various classes.

The second phase focuses on combining data from multiple interdependent networks to uncover more information that could be crucial for effective decision-making for AI systems. The work in this phase includes developing strategies for decreasing the computational demands on individual devices while combining multiple data sources for improved efficiency.

The final phase of the project will focus on privacy and security concerns and is aimed at preventing hacking or other intrusions.

“This not only strengthens the security of the model training but also ensures the protection of sensitive crime-related data,” Imteaj said. “This project could enhance the capabilities of law enforcement by providing them with more sophisticated and effective tools for proactive and informed decision-making.”

Imteaj previously received another grant aimed at using AI for law enforcement. That one-year, $100,000 grant from the U.S. Department of Homeland Security under its Criminal Investigations and Network Analysis Center program explored how criminals might use AI technology and how law enforcement might use it to detect and combat criminal behavior.

The state-of-the-art training he intends to create in this previous work is urgently needed, with AI becoming an ever more common presence in the world. His work will develop educational modules and certificate programs covering topics related to analytics, computer vision and integrating criminology perspectives into AI investigations.