I'm a second-year Ph.D. student at the University of Illinois Urbana-Champaign, advised by Xiaojing Liao.

My research focuses on making AI models more transparent and accountable. I'm interested in how we can verify what data models were trained on, what computations they actually perform, and what they are capable of doing. To address these questions, I bring together systems security, cryptography, and empirical measurement.

Previously, I received my M.S. in Computer Science from KAIST, where I worked on blockchain security with Min Suk Kang in the Network Security and Privacy Lab. I received my B.S. in Computer Science from Yonsei University.

Research interests

  • Verifiable AI systems: provable LLM inference, zkML, information-flow control, and GPU-accelerated proving
  • Training-data provenance: provable data membership inference
  • LLM evaluation and measurement: developing methods to evaluate benchmarks reliable

Publications

  1. Bits Under ZK-LLM: Evaluating Zero-Knowledge-Friendly Quantization for Verifiable Private LLM Inference

    Taeung Yoon, Yupeng Zhang, Xiaojing Liao

    Preprint · Under review

    A systematic study of how quantization precision affects LLM utility and the cost of zero-knowledge proofs.

  2. Privacy-Preserving Distributed Virtual Network Embedding

    Taeung Yoon, Lincoln Thurlow

    Cyber Security Experimentation and Test Workshop (CSET) 2026 · To appear

    Acceptance rate: 37%

    3pViNE uses zero-knowledge proofs to verify valid virtual network embeddings across multiple domains without revealing providers’ private network topologies.

  3. On Frontrunning Risks in Batch-Order Fair Systems for Blockchains

    Eunchan Park*, Taeung Yoon*, Hocheol Nam, Deepak Maram, Min Suk Kang

    ACM Conference on Computer and Communications Security (CCS) · October 2025

    * Equal contribution · Acceptance rate: 13.9%

    The Ambush attack exposes frontrunning risks in batch-order fair systems; delayed gossiping helps mitigate these risks while preserving throughput.

  4. Partitioning Ethereum without Eclipsing It

    Hwanjo Heo, Seungwon Woo, Taeung Yoon, Min Suk Kang, Seungwon Shin

    Network and Distributed System Security Symposium (NDSS) · February 2023

    Acceptance rate: 16.4%

    Gethlighting partitions Ethereum nodes without controlling every peer connection; the work also develops countermeasures.

News

  • Our paper “Privacy-Preserving Distributed Virtual Network Embedding” was accepted to CSET 2026!
  • Serving on the Program Committee of CSET 2026.
  • Our preprint “Bits Under ZK-LLM: Evaluating Zero-Knowledge-Friendly Quantization for Verifiable Private LLM Inference” is available on arXiv.
Earlier updates

Service & awards

Program Committee

Cyber Security Experimentation and Test (CSET)

2026

Undergraduate Research Apprenticeship Program Mentor

University of Illinois Urbana-Champaign

2026–2027

CS Graduate Ambassador

University of Illinois Urbana-Champaign

Mar 2026

Artifact Evaluation Committee

USENIX Security 2026 · NDSS 2027

2026

Travel Grant ($1,000)

ACM CCS

Oct 2025

2nd Place, Near Horizon AI Hackathon ($4,000)

Near protocol

Sep 2024

Teaching

Advanced Topics in Security, Privacy, and Machine Learning

Teaching Assistant · UIUC

Fall 2026

Introduction to Computer Networks

Teaching Assistant · KAIST

Fall 2024

Introduction to Information Security

Teaching Assistant · KAIST

Spring 2024