Hero Slide 1
Hero Slide 2
Hero Slide 3

Data Analysis & Machine Intelligence Lab

Department of Computer Science and Artificial Intelligence

DONGGUK UNIVERSITY

신뢰할 수 있는 AI와 AI·로봇 시스템의 안전성 및 보안을 연구합니다.

Trustworthy AI · AI Safety & Security · Embodied AI & Robot Safety

About Us

Welcome to Data Analysis & Machine Intelligence Lab at Dongguk University. Our lab aims at designing practical data science and machine learning systems that are ethically fair, privacy and security assured, algorithmically stable, and interpretable to humans.

Our current research centers on Trustworthy AI, AI Safety & Security, and Safe Embodied AI: the safety and security of large language models, AI agents, and robot systems.

P1

Research Directions

Trustworthy AI

AI가 언제, 왜 잘못된 결과를 내는지 먼저 밝히고, 그것을 막는 방법을 설계합니다. 대상은 챗봇 같은 대규모 언어모델, 스스로 도구를 쓰는 AI 에이전트, 그리고 로봇처럼 실제 환경에서 움직이는 시스템입니다.

01

AI Safety & Security

챗봇과 AI 에이전트를 먼저 공격해 위험한 답이 나오는 지점을 찾고 막습니다.

  • LLM Safety & Security
  • Jailbreaking & Adversarial Attacks
  • Agent Safety
02

Embodied AI & Robot Safety

로봇이 지시받은 일을 하다가 위험한 행동으로 넘어가지 않게 만듭니다.

  • Safety-aware Robot Planning
  • Safe Vision-Language-Action / VLM Systems
  • Physical AI Safety
  • Robot Security
03

Privacy & Unlearning

AI가 배운 것 중 지워야 할 정보만 골라 지우고, 데이터는 보호한 채 학습합니다.

  • Machine Unlearning
  • Privacy-Preserving Learning
04

Robustness & Fairness

눈에 띄지 않는 노이즈나 데이터 편향 때문에 판단이 틀어지는 문제를 줄입니다.

  • Adversarial Robustness
  • Optimization Stability
  • Fairness
View Research

Collaborators

History

2026

  1. EMNLP 2026 논문 5편 채택

    Confess What You Know: Forget-Set Misalignment with Model Knowledge in LLM Unlearning (Main)

    Will the User Ever Know? Covert Indirect Prompt Injection Attacks on Tool-Using LLM Agents (Main)

    FaVOR: LLM-Based Agentic Framework for Factor Mining via Empirical Validation (Findings)

    SafeBranch: Branch-Pair Safety Alignment for Embodied Agents (Findings)

    HiP-LR: Hierarchical Planning with Localized Failure Recovery for Long-Horizon Multi-Robot Tasks (Findings)

  2. 김미소, 김현진 학생, 석사과정 졸업

    이하늘 학생, 학사과정 졸업 후 석사과정 입학

  3. Research in International Business and Finance (JCR Top 4%) 논문 게재

    Who provides liquidity in retail-dominated markets? Evidence from Korea

    SSCI, JCR Q1, IF: 7.6, JCR Top 4% (9/243)

  4. 김민준, 장윤영, 조찬웅 학생 입학

    학석사연계과정 김민준·장윤영, 인턴 조찬웅

  5. 김미소 연구원, 네이버 클라우드 인턴 합격

  6. Journal of Hazardous Materials Advances (JCR Top 9%) 논문 게재

    Sampling-bias-aware machine-learning mapping of geogenic indoor radon hazard in Jeollanam-do, Korea using measured-low negatives

    SCIE, JCR Q1, IF: 9.1, JCR Top 9%

  7. 김현진 연구원, HD한국조선해양 생성형 언어모델 연구직 합격

  8. Geosciences Journal (JCR Q2) 논문 게재

    Groundwater potential mapping using a frequency ratio-LSTM hybrid model in the Andijan Region, Uzbekistan

    SCIE, JCR Q2, IF: 1.5

  9. ICML 2026 워크샵 논문 채택

    What Did You Do Behind My Back?! Covert Indirect Prompt Injection on Tool-Using LLM Agents (FAGEN@ICML 2026)

  10. IEEE Access (JCR Q2) 논문 게재

    Exploring the Effect of Multi-Step Ascent in Sharpness-Aware Minimization

    SCIE, JCR Q2, IF: 3.6

  11. Applied Sciences (JCR Q2) 논문 게재

    Explainable Graph-Based Golf Swing Analysis Integrating Club and Body Keypoints for Ball Flight Outcome Prediction

    SCIE, JCR Q2, IF: 2.5

  12. 정승현 학생, 박사과정 입학

  13. Journal of Theoretical and Applied Electronic Commerce Research (JCR Q1) 논문 게재

    Disconfirmation Dynamics in Service Recovery: Insights from Online Customer Reviews

    SSCI, JCR Q1, IF: 4.5

  14. ICLR 2026 논문 2편 채택

    Co-occurring Associated REtained concepts in Diffusion Unlearning (Poster)

    SlotGCG: Exploiting the Positional Vulnerability in LLMs for Jailbreak Attacks (Poster)

2025

  1. Computational Economics (JCR Q2) 논문 게재

    Large Language Models as Financial Analysts: Sector-Aware Reasoning for Investment Decisions

    SCIE, JCR Q2, IF: 2.2

  2. 졸업생 Christoph Timmermann 학생, Technical University of Munich × Audi 박사과정 합격

  3. 정승현 학생, 한국정보처리학회 ACK 2025 산학연협동우수상 수상

  4. 김캐디 산학 과제 수행

    골프 클럽과 관절 정보를 활용한 골프 스윙 분석 (2025.09 - 2026.01)

  5. 이윤석, 김윤지, 정승원, 정지우 학생, 석사과정 입학

  6. 이현세 학생, 한국데이터마이닝학회 2025 하계학술대회 최우수 논문상 수상

  7. Geomatics, Natural Hazards and Risk (JCR Top 15%) 논문 게재

    Deep learning-enhanced geospatial modeling for indoor radon mapping in Jeollabuk-do, South Korea

    SCIE, JCR Q1, IF: 4.4, JCR Top 15%

  8. CVsports at CVPR 2025 논문 게재

    CaddieSet: A Golf Swing Dataset with Human Joint Features and Ball Information

  9. 김윤지, 이윤석 학생 입학

    학석사연계과정 김윤지, 학부 인턴 이윤석

  10. Information Sciences (JCR Top 8%) 논문 게재

    Towards undetectable adversarial attack on time series classification

    SCIE, JCR Q1, IF: 8.2, JCR Top 8%

  11. IEEE Access (JCR Q2) 논문 게재

    LM-CLIP: Adapting Positive Asymmetric Loss for Long-Tailed Multi-Label Classification

    SCIE, JCR Q2, IF: 3.6

  12. Finance Research Letters (JCR Top 1%) 논문 게재

    Interconnected dynamics of sustainable cryptocurrencies: Insights from transfer entropy analysis

    SSCI, JCR Q1, IF: 7.4, JCR Top 1%

  13. 기상청 산학 과제 수행

    전천(All sky) 기후자료 산출 및 초해상화 인공지능 모델 개발 (2025.04 - 2027.12)

  14. 한국과학기술정보연구원 산학 과제 수행

    지능형 데이터 큐레이션 및 통합 구축 기술 개발 (2025.04 - 2025.09)

  15. 한국연구재단 우수신진연구 선정

    윤리적인 다국어 LLM 활용을 위한 문맥 기반 탈옥 가드레일 구축 (2025.03 - 2027.02)

  16. 이현세 학생, 석사과정 입학

  17. Applied Soft Computing (JCR Top 10%) 논문 게재

    Black-box adversarial examples via frequency distortion against fault diagnosis systems

    SCIE, JCR Q1, IF: 7.2, JCR Top 10%

  18. 김민석 학생, 석사과정 입학

  19. Earth Systems and Environment (JCR Top 10%) 논문 게재

    AI-Driven Geospatial Analysis of Indoor Radon Levels: A Case Study in Chungcheongbuk-do, South Korea

    ESCI, JCR Q1, IF: 5.3, JCR Top 10%

2024

  1. Expert Systems with Applications (JCR Top 10%) 논문 게재

    Multi-stage ensemble with refinement for noisy labeled data classification

    SCIE, JCR Q1, IF: 7.5, JCR Top 10%

  2. 사물인터넷 혁신융합대학사업 산학 과제 수행

    예외 상황과 일반 상황의 분포 차이 최소화 적응 기법 개발 (2024.09 - 2025.01)

  3. 김현진, 김미소 학생, 석사과정 입학

  4. 김민석 학생, 학석사연계과정 입학

  5. Journal of Environmental Management (JCR Top 10%) 논문 게재

    Flood susceptibility mapping of Cheongju, South Korea based on the integration of environmental factors using various machine learning approaches

    SCIE, JCR Q1, IF: 8.0, JCR Top 10%

  6. Journal of Ambient Intelligence and Humanized Computing (JCR Top 10%) 논문 게재

    Sliced Wasserstein adversarial training for improving adversarial robustness

    SCIE, JCR Q1, IF: 7.104, JCR Top 10%

  7. Pattern Recognition Letters (JCR Q2) 논문 게재

    Graph contrastive learning with consistency regularization

    SCIE, JCR Q2, IF: 3.9

  8. Engineering Applications of Artificial Intelligence (JCR Top 2%) 논문 게재

    Evaluating practical adversarial robustness of fault diagnosis systems via spectrogram-aware ensemble method

    SCIE, JCR Q1, IF: 7.5, JCR Top 2%

  9. 이거루, 정승현 학생 대학원 입학

    석박사통합과정 이거루, 석사과정 정승현

  10. 김캐디 산학 과제 수행

    XAI를 활용한 골프 스윙 피드백 모델 개발 (2024.01 -

2023

  1. IEEE Access (JCR Q2) 논문 게재

    Fairness-Aware Multimodal Learning in Automatic Video Interview Assessment

    SCIE, JCR Q2, IF: 3.9

  2. Neural Networks (JCR Top 11%) 논문 게재

    Bridged adversarial training

    SCIE, JCR Q1, IF: 6.6, JCR Top 11%

  3. Applied Soft Computing (JCR Top 10%) 논문 게재

    Fast sharpness-aware training for periodic time series classification and forecasting

    SCIE, JCR Q1, IF: 7.2, JCR Top 10%

  4. Christoph Timmermann 학생, 석사과정 입학

  5. Scientific Reports (JCR Q2) 논문 게재

    Genetic descriptor search algorithm for predicting hydrogen adsorption free energy of 2D material

    SCIE, JCR Q2, IF: 4.6

  6. 정승원, 정지우, 김미소, 김현진 학생, 학석사연계과정 입학

  7. Pixt 산학 과제 수행

    한국어 전용 동영상 캡션 AI 모델 개발 (2023.05 - 2024.04)

2022

  1. Pattern Recognition (JCR Top 7%) 논문 게재

    Variational cycle-consistent imputation adversarial networks for general missing patterns

    SCIE, JCR Q1, IF: 7.5, JCR Top 7%

  2. 정승현 학생, 학석사연계과정 입학

  3. 제네시스랩 산학 과제 수행

    공정한 인공지능 면접 모델 알고리즘 개발 (2022.07 - 2022.12)

  4. 한국연구재단 산학 과제 수행

    인공지능 시스템의 실용화를 위한 딥러닝 안정성 알고리즘 개발 (2022.06 - 2025.02)

2021

  1. NeurIPS 논문 게재

    Parameter-free HE-friendly Logistic Regression

  2. NeurIPS 논문 게재

    Towards Better Understanding of Training Certifiably Robust Models against Adversarial Examples

  3. Information Sciences (JCR Top 10%) 논문 게재

    Fair clustering with fair correspondence distribution

    SCIE, JCR Q1, IF: 8.233, JCR Top 10%

  4. Pattern Recognition (JCR Top 7%) 논문 게재

    Compact class-conditional domain invariant learning for multi-class domain adaptation

    SCIE, JCR Q1, IF: 7.5, JCR Top 7%