産総研：川本裕輔

Profile

I am the leader of Continuum Computing Trustworthiness Research Group at AIST (National Institute of Advanced Industrial Science and Technology).

Research

My research topic is the foundation of information security and software science, especially the modeling and verification of probabilistic programs, protocols, and systems by using the following approaches and their integration:

symbolic/logical approach (e.g., formal methods, programming languages, mathematical logic);
probabilistic/statistical approach (e.g., machine learning, statistics, information theory).

My recent research includes (i) specification languages and automated verification techniques for the correct usage of statistics, (ii) symbolic logic for statistical causality, (iii) risk analyses and formal verification of AI systems, and (iv) software verification techniques for the AI era.

Examples of my research interests:

Formal specification, verification, and quality management of statistics, machine learning, and AI systems

Logic and automated verification for the correct usage of statistics [CAV25] [JELIA23] [AI23] [KR21]
Epistemic logic and possible world semantics for machine learning [SoSyM20]
Taxonomy for machine learning security [preprint23a]
Quality management guidelines for machine learning systems [AIQM Guidelines]

Formal methods for probabilistic systems and quantitative properties
- Statistical approach to program analysis [FAOC19] [ATVA17] [FM16] [ESORICS14] [CAV13]
- Foundation of quantitative information flow (QIF) [LMCS17] [QAPL15] [CSF13]
Foundation of privacy protection mechanisms
- Game theory for information leakage (QIF/DP) [ACM TOPS22] [Entropy18]
- Extension of differential privacy (DP) [ESORICS21] [Allerton19] [ESORICS19] [USENIX Sec19]

Here is a publication list. Details can be found in researchmap [Japanese | English]

Our group

Our research group (Group Leader: Dr. Yusuke Kawamoto) is working on the following topics:

development and social implementation of evaluation criteria for AI systems
research on the quality evaluation and management for AI systems
research on the formal methods for AI systems, cyber-physical systems, AI-generated software, etc.
standardization and social implementations related to digital architecture.

Open positions

Master's and Ph.D. students at NAIST (Nara Institute of Science and Technology), which is a national university different from AIST.
Research assistant positions at AIST (for master's and Ph.D. students studying at other universities in Japan)
Post-doctoral researcher positions at AIST (AIST postdocs and JSPS postdocs)
Young permanent researcher positions at AIST (for AI safety and for formal methods)

For inquiries regarding these positions, please contact us at M-ipri-cctrg-inquery-ml at aist.go.jp

Previously

I was working on the following research topics:

Formal methods for cryptographic protocols and their computational soundness [CompSoft16] [AMH13] [ISPEC12]
Attacks on RSA PKCS#1 v1.5 on cryptographic device [CRYPTO12]
Attacks on the privacy of deep transfer learning [IJCNN21]