Organization
School of Computing Professor
Title
Professor
Contact information
External link
SAKUMA JUN
Research Areas
-
Informatics / Intelligent informatics
Research History
-
Institute of Science Tokyo School of Computing Professor
2023.4
Professional Memberships
-
The Institute of Electronics, Information and Communication Enineers (IEICE)
2012.4
-
THE JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE
2004.3
Committee Memberships
-
ISO SC27/WG5 (アイデンティティ管理とプライバシー技術) 専門委員会 専門委員会委員
2016.4
Papers
-
One Leak Away: How Pretrained Model Exposure Amplifies Jailbreak Risks in Finetuned LLMs. Reviewed
Yixin Tan, Zhe Yu, Jun Sakuma
CoRR abs/2512.14751 2025.12
-
Disrupting Model Merging: A Parameter-Level Defense without Sacrificing Accuracy. Reviewed International journal
Junhao Wei, Yu Zhe, Jun Sakuma
ICCV 17698 - 17707 2025.10
-
Differentially Private Synthetic Text Generation for Retrieval-Augmented Generation (RAG). Reviewed
Junki Mori, Kazuya Kakizaki, Taiki Miyagawa, Jun Sakuma
CoRR abs/2510.06719 2025.10
-
Ragib Amin Nihal, Rui Wen, Kazuhiro Nakadai, Jun Sakuma
CoRR abs/2510.08859 2025.10
-
Model-agnostic Adversarial Attack and Defense for Vision-Language-Action Models. Reviewed
Haochuan Xu, Yun Sing Koh, Shuhuai Huang, Zirun Zhou, Di Wang 0015, Jun Sakuma, Jingfeng Zhang
CoRR abs/2510.13237 2025.10
-
Toward Safer Diffusion Language Models: Discovery and Mitigation of Priming Vulnerability. Reviewed
Shojiro Yamabe, Jun Sakuma
CoRR abs/2510.00565 2025.10
-
Explainable Classifier for Malignant Lymphoma Subtyping via Cell Graph and Image Fusion. Reviewed International journal
Daiki Nishiyama, Hiroaki Miyoshi, Noriaki Hashimoto, Koichi Ohshima, Hidekata Hontani, Ichiro Takeuchi, Jun Sakuma
Medical Image Computing and Computer Assisted Intervention - MICCAI 2025 - 28th International Conference 320 - 330 2025.9
-
Weakening Prediction Confidence Makes Backdoors Strengthened: A Strong Textual Backdoor Attack on Prefix-tuning. Reviewed International coauthorship International journal
Yixin Tan, Haoyu Zhang, Jun Sakuma
IJCNN 1 - 8 2025.6
-
Remembering Transformer for Continual Learning. Reviewed International journal
Yuwei Sun, Ippei Fujisawa, Arthur Juliani, Jun Sakuma, Ryota Kanai
IJCNN 1 - 9 2025.6
-
Harnessing the Power of Vicinity-Informed Analysis for Classification under Covariate Shift. Reviewed International journal
Mitsuhiro Fujikawa, Youhei Akimoto, Jun Sakuma, Kazuto Fukuchi
International Conference on Artificial Intelligence and Statistics(AISTATS) 226 - 234 2025.3
-
Explainable Classifier for Malignant Lymphoma Subtyping via Cell Graph and Image Fusion. Reviewed
Daiki Nishiyama, Hiroaki Miyoshi, Noriaki Hashimoto, Koichi Ohshima, Hidekata Hontani, Ichiro Takeuchi, Jun Sakuma
CoRR abs/2503.00925 2025.3
-
Data Poisoning Attacks to Locally Differentially Private Range Query Protocols. Reviewed
Ting-Wei Liao, Chih-Hsun Lin, Yu-Lin Tsai, Takao Murakami, Chia-Mu Yu, Jun Sakuma, Chun-Ying Huang, Hiroaki Kikuchi
CoRR abs/2503.03454 2025.3
-
Disrupting Model Merging: A Parameter-Level Defense Without Sacrificing Accuracy. Reviewed
Junhao Wei, Yu Zhe, Jun Sakuma
CoRR abs/2503.07661 2025.3
-
Deterministic and Probabilistic Certified Defenses for Content-Based Image Retrieval. Reviewed International journal
Kazuya Kakizaki, Kazuto Fukuchi, Jun Sakuma
IEICE Trans. Inf. Syst. 108 ( 1 ) 92 - 109 2025.1
-
BADTV: Unveiling Backdoor Threats in Third-Party Task Vectors. Reviewed
Chia-Yi Hsu, Yu-Lin Tsai, Yu Zhe, Yan-Lun Chen, Chih-Hsun Lin, Chia-Mu Yu, Yang Zhang, Chun-Ying Huang, Jun Sakuma
CoRR abs/2501.02373 2025.1
-
Trojan attribute inference attack on gradient boosting decision trees. Reviewed International journal
Kunihiro Ito, Batnyam Enkhtaivan, Isamu Teranishi, Jun Sakuma
9th IEEE European Symposium on Security and Privacy(EuroS&P) 542 - 559 2024.7
-
Linear Convergence Rate Analysis of the (1+1)-ES on Locally Strongly Convex and Lipschitz Smooth Functions. Reviewed International journal
Daiki Morinaga, Kazuto Fukuchi, Jun Sakuma, Youhei Akimoto
Proceedings of the Genetic and Evolutionary Computation Conference Companion 49 - 50 2024.7
-
Instance-Level Trojan Attacks on Visual Question Answering via Adversarial Learning in Neuron Activation Space. Reviewed International journal
Yuwei Sun, Hideya Ochiai, Jun Sakuma
International Joint Conference on Neural Networks(IJCNN) 1 - 9 2024.7
-
Convergence Rate of the (1+1)-ES on Locally Strongly Convex and Lipschitz Smooth Functions. Reviewed International journal
Daiki Morinaga, Kazuto Fukuchi, Jun Sakuma, Youhei Akimoto
IEEE Transactions on Evolutionary Computation 28 ( 2 ) 501 - 515 2024.4
-
Attacking-Distance-Aware Attack: Semi-targeted Model Poisoning on Federated Learning. Reviewed International journal
Yuwei Sun, Hideya Ochiai, Jun Sakuma
IEEE Transactions on Artificial Intelligence 5 ( 2 ) 925 - 939 2024.2
-
Zero-shot domain adaptation based on dual-level mix and contrast. Reviewed
Yu Zhe, Jun Sakuma
CoRR abs/2406.18996 2024
-
Parameter Matching Attack: Enhancing Practical Applicability of Availability Attacks. Reviewed
Yu Zhe, Jun Sakuma
CoRR abs/2407.02437 2024
-
Remembering Transformer for Continual Learning. Reviewed
Yuwei Sun, Ippei Fujisawa, Arthur Juliani, Jun Sakuma, Ryota Kanai
CoRR abs/2404.07518 2024
-
Adversarial Attacks on Hidden Tasks in Multi-Task Learning. Reviewed
Yu Zhe, Rei Nagaike, Daiki Nishiyama, Kazuto Fukuchi, Jun Sakuma
CoRR abs/2405.15244 2024
-
Harnessing the Power of Vicinity-Informed Analysis for Classification under Covariate Shift. Reviewed
Mitsuhiro Fujikawa, Yohei Akimoto, Jun Sakuma, Kazuto Fukuchi
CoRR abs/2405.16906 2024
-
Behavior-Targeted Attack on Reinforcement Learning with Limited Access to Victim's Policy. Reviewed
Shojiro Yamabe, Kazuto Fukuchi, Ryoma Senda, Jun Sakuma
CoRR abs/2406.03862 2024
-
Zero-shot domain adaptation based on dual-level mix and contrast. Reviewed
Yu Zhe, Jun Sakuma
IEEE Conference on Artificial Intelligence(CAI) 1549 - 1556 2024
MISC
-
Targeted Manipulation Attacks on Reinforcement Learning Agents through Imitation Learning with Perturbed Observations
山辺翔二郎, 福地一斗, 福地一斗, 仙田涼摩, 佐久間淳, 佐久間淳
人工知能学会全国大会論文集(Web) 38th 2024
Presentations
-
Prediction with Model-based Neutralization International conference
Jun, Sakuma, Kazuto, Fukuchi, Toshihiro, Kamishima
The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) 2013 2013.9
-
Privacy-Preserving Protocol for Epidemiology in Effect of Radiation International conference
佐久間,淳, Hiroaki,Kikuchi, Tomoki,Sato
The Seventh International Conference on Complex, Intelligent, and Software Intensive Systems (CISIS-2013) 2013.7
-
Considerations on Fairness-aware / Discrimination-aware Mining International conference
佐久間,淳, Toshihiro,Kamishima, Shotaro,Akaho
The International Workshop on Discrimination and Privacy-Aware Data Mining (DPADM), ICDM2012 2012.12
-
Applicability of Existing Anonymization Methods to Large Location History Data in Urban Travel
繁富理恵ほか, 佐久間,淳
2012 IEEE International Conference on Systems, Man, and Cybernetics
-
Bloom Filter Bootstrapping: Privacy-Preserving Estimation of the Size of an Intersection
佐久間,淳
27th Annual IFIP WG 11.3 Working Conference on Data and Applications Security and Privacy (DBSec '13)
-
Ice and Fire: Quantifying the Risk of Re-identification and Utility in Data Anonymization International conference
Sakuma, Jun, Kikuchi.Hiroaki, Yamaguchi, Takayasu, Hamada, Koki, Yamaoka, Yuji, Oguri, Hidenobu
The 30th IEEE International Conference on Advanced Information Networking and Applications
-
Link Analysis for Private Weighted Graphs International conference
Sakuma, Jun, Kobayashi, Shigenobu
32nd Annual International ACM SIGIR Conference on Research and Development in Information Retrieval
-
Privacy-Preserving Evaluation of Generalization Error and Its Application to Model and Attribute Selection International conference
Sakuma, Jun, Wright, Rebecca N
1st Asian Conference on Machine Learning
-
Fairness-aware Classifier with Prejudice Remover Regularizer International conference
佐久間,淳
The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) 2012
-
Enhancement of the Neutrality in Recommendation
佐久間,淳
6th ACM Recommender Systems 2012, Workshop on Human Decision Making in Recommender Systems
Research Projects
-
合成データ生成によるロバストな連合基盤モデル
Grant number：25126212 2025 - 2027
科学技術振興機構 国際的な科学技術共同研究などの推進/日ASEAN科学技術・イノベーション協働連携事業（NEXUS）/国際共同研究
佐久間 淳, 村上 隆夫
本研究は、安全かつ効率的な基盤モデルの学習を、データを分散させたまま学習させる連合学習を通じて実現するためのフレームワークの構築を目指す。
具体的には、日本側チームは、モデル学習におけるプライバシーの問題及び分散したモデルの統合におけるセキュリティの問題に取り組み、シンガポール側チームは、信頼される基盤モデルを連合学習によって実現するための方法論開発とそのセキュリティ・プライバシー保護に取り組む。
両国の研究チームによる共同研究を通じて、基盤モデルの連合学習において懸念されるセキュリティやプライバシーの問題解決が期待される。 -
大規模言語モデルのミスアライメントに対するレッドチーミング基盤
Grant number：24017354 2024 - 2029
科学技術振興機構 戦略的な研究開発の推進/経済安全保障重要技術育成プログラム
佐久間 淳
本研究開発は、大規模生成モデルの生成コンテンツのミスアライメント（人間の期待や倫理観から外れた挙動）を検出し、これを軽減・抑制するためのセキュリティ技術基盤の構築を目的としています。大規模生成モデルの利用においては、その生成コンテンツに有害情報・偽情報・差別的内容が含まれていたり、その生成コンテンツによって機密漏えい、プライバシー侵害、著作権侵害などが発生したりする可能性があり、このようなミスアライメントへの対策が不可欠です。一般の開発者が外部から入手した大規模生成モデルをそのまま利用したり、これを手元のデータで改変して利用したりする状況では、このようなミスアライメントのリスクを独力で評価することは簡単ではありません。本研究開発では、生成モデルのこのようなミスアライメントに関するリスクの評価を支援するための技術基盤を提供し、ミスアライメントの抑制につなげます。
-
Understanding Attack Mechanisms against AI through Causal Structures of Classification and Building Countermeasures
Grant number：23H00483 2023.4 - 2028.3
Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research Grant-in-Aid for Scientific Research (A)
Grant amount：\46800000 （ Direct Cost: \36000000 、 Indirect Cost：\10800000 ）
-
分散型ソーシャルグラフに向けた差分プライバシー技術
Grant number：22H00521 2022.4 - 2027.3
日本学術振興会 科学研究費助成事業 基盤研究(A)
村上 隆夫, 南 和宏, 日野 英逸, Attrapadung Nuttapong, 曹 洋, 大原 一真, 佐久間 淳
Grant amount：\41860000 （ Direct Cost: \32200000 、 Indirect Cost：\9660000 ）
-
Understanding the impact of adversarial inputs on autonomous driving systems and developing countermeasure technologies
Grant number：22H00519 2022.4 - 2025.3
Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research Grant-in-Aid for Scientific Research (A)
Grant amount：\42380000 （ Direct Cost: \32600000 、 Indirect Cost：\9780000 ）