Skip to main content
Alumnus homeNews home
Story
23 of 50

LSU Computer Science's Baggili Working to Address Artificial Intelligence Investigations

September 25, 2024BATON ROUGE, LA - Artificial intelligence (AI) is widely used on a daily basis, whether it's asking Alexa or nighttime with Netflix. But what happens if there are failures within the AI, and how are they investigated properly? For instance, if a hospital created an AI model that could detect a medical diagnosis and it led to an unnecessary treatment, that model would need to be investigated for how it was trained, who trained it, and on what machine it was trained.

Abe Baggili and Clinton WalkerResearch Made Possible Through New $400k NSF Grant

September 25, 2024

BATON ROUGE, LA – Artificial intelligence (AI) is widely used on a daily basis, whether it’s asking Alexa or nighttime with Netflix. But what happens if there are failures within the AI, and how are they investigated properly? For instance, if a hospital created an AI model that could detect a medical diagnosis and it led to an unnecessary treatment, that model would need to be investigated for how it was trained, who trained it, and on what machine it was trained.

Such investigations are at the heart of the work done by LSU Computer Science Chair and Professor Ibrahim “Abe” Baggili, who recently received a competitive $400,000 grant from the National Science Foundation’s SaTC CORE program to advance digital forensics techniques around AI. Digital forensics involves the scientific acquisition, authentication, and analysis of digital evidence as it applies to the law.

Working with Baggili on this project are his PhD student, Clinton Walker, a native of Monroe, La.; and Hao Wang, assistant professor in the Department of Electrical and Computer Engineering at the Stevens Institute of Technology in Hoboken, N.J.

“To get evidence admitted to the court of law, we have to use techniques that are scientifically valid, published, and tested,” Baggili said. “At the moment, no such tools exist in the domain of model forensics. Therefore, we would be helping the community develop those tools and techniques and have the necessary science behind them so that investigators in the future have the ability to perform such investigations.

“Back in the day, there was no way to place someone at the crime scene without science, until blood types were identified. Blood types aren’t great, of course, because people share blood types, so we moved to DNA to then be able to specifically identify an individual. We are hoping we move closer to that with our scientific discoveries.”

Indeed, a successful outcome for this project would have a significant impact on the cybersecurity industry and far-reaching implications for industries that rely heavily on machine learning models—i.e., government, finance, healthcare, and transportation, for example—by improving their ability to detect and mitigate potential threats and vulnerabilities.

“AI has infiltrated our lives and is being used behind the scenes when we buy things online, with our banking, social media, etc.,” Baggili said. “We cannot escape it. Hopefully, we can figure out how to investigate AI failures before it’s too late. While AI is really cool and continues to grow rapidly as a domain and its impact on us, little to no work is really being focused on how to investigate a failure when it happens. We are really excited to pioneer in this area.”

Like us on Facebook (@lsuengineering) or follow us on X (formerly Twitter) and Instagram (@lsuengineering).​

###

Contact: Joshua Duplechain
Director of Communications
225-578-5706
josh@lsu.edu

Latest College of Engineering News