Security and Robustness of Efficient AI Models
Studying bit-flip and backdoor attacks on efficient neural networks and long-tail dataset robustness.
Duration: Jan. 2024 – Dec. 2025 Advisor: Dr. Rui Ning, Old Dominion University
Studied bit-flip and backdoor attacks on efficient neural networks. Observed attack success rate (ASR) saturation at 60–70% on imbalanced long-tail datasets, providing insights into the inherent robustness properties of data distribution.
Keywords: Model Robustness, Adversarial ML, Experimental Evaluation