Federal University of Amazonas (UFAM) Seminar: AICodeRepair
Back
Previous Article: Favourite LaTeX Templates
Next Article: Formal Methods of Language Modelling
Author: Yiannis Charalambous
Reading Time: 1 minute
Tags: AI · Security
I recently presented our group’s latest research to the Federal University of Amazonas (UFAM). The experiments were conducted from January 2024 and April 2024. The video can be watched below.
Abstract
As society increasingly depends on artificial intelligence, the need for trust in neural networks becomes more critical. However, neural networks often exhibit vulnerabilities that can undermine this trust. In this seminar, I will present our research group’s efforts to enhance the reliability of neural networks embedded in C code. We utilize Large Language Models in conjunction with Bounded Model Checking to identify and repair these vulnerabilities effectively.
Acknowledgements
Thank you to the Federal University of Amazonas (UFAM) and Iury Bessa for inviting me to present our work. Many thanks to my Supervisor Lucas C. Cordeiro and Edoardo Manino for helping me conduct this research.