IEEE EuroS&P 2026
SecDTD: Dynamic Token Drop for Secure Transformers Inference
Yifei Cai, Zhuoran Li, Yizhou Feng, Qiao Zhang, Hongyi Wu, Danella Zhao, and Chunsheng Xin.
AI Security & Privacy · Efficient Secure Inference · Privacy-Preserving AI Systems
Yifei Cai
Postdoctoral Research Associate · Iowa State University
I develop practical privacy-preserving AI systems by co-designing machine learning models and secure computation—from HE-aware model optimization to efficient Transformer inference and selective protection for LLMs.
About
I am a Postdoctoral Research Associate in the Department of Computer Science at Iowa State University. My research has progressed from cryptography-aware model pruning and secure-inference-oriented architectures to efficient secure Transformer inference. My current work investigates selective privacy protection for LLMs and practical, reusable secure AI systems. I also mentor Ph.D. students working in AI security and privacy.
Research
Research overview →Privacy-Preserving ML
Cryptography-aware structured pruning and neural architecture design that reduce the cost of homomorphic-encryption-based inference.
Secure & Efficient AI Systems
Accelerating secure Transformer inference through dynamic token dropping, oblivious selection, and model–protocol co-design.
Secure & Private Foundation Models
Exploring selective privacy protection for LLMs while balancing privacy requirements, model utility, and execution cost.
Selected Publications
All publications →IEEE SaTML 2026
Towards Zero Rotation and Beyond: Architecting Neural Networks for Fast Secure Inference with Homomorphic Encryption
Yifei Cai, Yizhou Feng, Qiao Zhang, Chunsheng Xin, and Hongyi Wu.
ICML 2026
DF-LoGiT: Data-Free Logic-Gated Backdoor Attacks in Vision Transformers
Xiaozuo Shen, Yifei Cai, Rui Ning, Chunsheng Xin, and Hongyi Wu.
ACM AsiaCCS 2024
MOSAIC: A Prune-and-Assemble Approach for Efficient Model Pruning in Privacy-Preserving Deep Learning
Yifei Cai, Qiao Zhang, Rui Ning, Chunsheng Xin, and Hongyi Wu.
ACM AsiaCCS 2022
Hunter: HE-Friendly Structured Pruning for Efficient Privacy-Preserving Deep Learning
Yifei Cai, Qiao Zhang, Rui Ning, Chunsheng Xin, and Hongyi Wu.
Teaching & Mentoring
Teaching & mentoring →I supported undergraduate and graduate cybersecurity courses at Old Dominion University through supplemental instruction, hands-on labs, and attack-and-defense exercises. In my current postdoctoral role, I mentor Ph.D. students in AI security and privacy, from research problem formulation to experimental evaluation. My teaching interests include AI, machine learning, cybersecurity, and privacy-preserving computation.