Biography
Dr. Mohammed Abuhamad is an Assistant Professor in the Department of Computer Science at Loyola University Chicago. He is the director and founder of the AI for Secure Computing Research Lab (AISeC) and serves as a faculty member in multiple labs, including Cybersecurity Lab and Software and Systems Lab. Dr. Abuhamad’s research focuses on AI system security, information security and privacy, responsible AI, and data analytics. He holds two Ph.D. degrees in Computer Science from the University of Central Florida and in Electrical and Computer Engineering from INHA University, South Korea. His work has been published in top-tier journals and conferences, and he is an active reviewer and committee member for several prestigious journals and conferences.
Talk: “Towards Enhancing Trust and Security in Interpretable Deep Learning Systems”
Abstract: Interpretation models have become an important aspect of deep learning development, providing the transparency and trust needed to detect data artifacts and refine problem formulations; however, this transparency also introduces a new, often overlooked attack surface. This talk provides a technical discussion of the security of interpretable deep learning systems by presenting various adversarial attacks designed to simultaneously deceive model predictions and their corresponding interpretations. We will analyze the effectiveness of these attacks, the resilience of standard defense mechanisms, and the potential for specific countermeasures to protect the interpretability layer. By highlighting the critical vulnerabilities in current frameworks, this talk offers practical strategies to improve attack circumvention and enhance the overall security of sensitive deep learning applications.