绿奴天花板

Riley Rutligliano 鈥�26 hopes people take away a sense of awareness from her Honors thesis, which she will present on Friday at the Honors Student Showcase.

Her thesis, 鈥淓valuating the Robustness of Deepfake Detection Models Under Real-World Distortions,鈥� uses a computer model to detect fake pictures, or artificial intelligence-generated photos, commonly referred to as deepfakes.

Rutligliano used that software to test 200 real and 200 fake photos to determine how well the model could detect AI distortions on photos.听

She tested cropping and downscaling to see how alterations of photos could impact the model. She then calculated AUROC, which is a metric used to determine how well a model can designate something into two groups 鈥斕齣n this case, real or fake.听

鈥淚鈥檝e always been intrigued with AI, and it鈥檚 emerging more and more every day,鈥� Rutligliano said. 鈥淎I鈥檚 not going anywhere, and we can鈥檛 fight it, so I鈥檇 rather do research into models we already have and see how well they鈥檙e being used to detect AI on photos.鈥�

The goal, she said, is to help people realize how deceiving AI can be, as the pictures don鈥檛 look fake, and it鈥檚 hard to distinguish by the naked eye.

鈥淏y doing this project, I think it could be a steppingstone into getting more people aware, and maybe looking at different patterns that people can see,鈥� she said.听

Following graduation, Rutligliano will be in 绿奴天花板ampa鈥檚 4+1 program to obtain a master鈥檚 degree in cybersecurity.

Her thesis is one of many that will be on display at the Honors Student Showcase from 1-3 p.m., Friday, in Plant Hall鈥檚 Fletcher Lounge.