計算機基礎学研究室は，広島大学大学院先進理工系科学研究科先進理工系科学専攻情報科学プログラムに属する研究室の1つで、現在次のような研究を行っています。

暗号技術と情報セキュリティの研究

テーマ：プライバシーを保護した認証
現在のユーザ認証では、IDの紐付けにより、サーバがユーザのアクセス履歴を把握できてしまいます。本グループでは、匿名性をもつユーザ認証についての研究を行なっています。正規利用者であるかを確認したい、年齢が20歳以上かを確認したいなど、IDを特定せずに認証のできる状況は多々あります。このような場合に、IDを秘匿しつつ属性など示したい情報のみ確認する認証について研究しています。
テーマ：ブロックチェーンにおけるプライバシー保護
分散型台帳であるブロックチェーンでは，各取引がオープンになるため，利用者のプライバシー保護が重要となります．本研究グループでは，プライバシーを保護できる匿名認証をブロックチェーンに応用する研究を進めています．

モバイルコンピューティング

AIプライバシーとセキュリティ

テーマ：連合学習における効率的なモデル集約と攻撃耐性の研究
連合学習（Federated Learning）は、データを共有せずにAIを学習できる技術として注目されています。本研究では、通信コストを抑えつつ精度を維持するモデル集約手法と、バックドア攻撃・モデル窃取攻撃への耐性向上を目的としたアルゴリズムの開発を行います。
テーマ：LLMに対する脱獄攻撃と防御手法の研究
大規模言語モデル（LLM）は、高度な対話能力を持つ一方で、脱獄攻撃（Jailbreak Attacks）により意図しない情報を生成するリスクがあります。本研究では、攻撃手法の分析を行い、安全性を向上させる防御技術を開発します。

The Foundation of Computer Science Laboratory is one of the laboratories belonging to the Information Science Program of the Department of Advanced Science and Engineering, Graduate School of Advanced Science and Engineering, Hiroshima University, and is currently conducting the following research.

Cryptography and Information Security

Topic: Privacy-Preserving Authentication
In conventional user authentication systems, user IDs are often linked to service access records, allowing servers to track user activity histories. Our research focuses on anonymous authentication mechanisms that enable verification without revealing user identity. In many practical scenarios, it is sufficient to confirm specific attributes—such as whether a user is authorized or over a certain age—without identifying the individual. We study authentication schemes that protect user identity while selectively disclosing only necessary attributes.
Topic: Privacy Protection in Blockchain Systems
Blockchain systems, as distributed ledgers, record transactions in an open and transparent manner, making privacy protection a critical concern. Our research explores the application of privacy-preserving anonymous authentication mechanisms within blockchain environments. We aim to design protocols that maintain the integrity and transparency of blockchain systems while safeguarding user anonymity and sensitive information.

モバイルコンピューティング

Topic: Automatic Identification of Familiar Devices and OS-Independent Direct File Transfer
This research explores technologies that enable mobile devices to simulate the process of human familiarity formation through repeated encounters. We study mechanisms for automatically identifying frequently encountered devices and for enabling secure, direct file transfer in close proximity without OS dependency. The work builds upon security protocols and communication technologies such as Bluetooth Low Energy (BLE), Wi-Fi, and peer-to-peer (P2P) networking.
Topic: Mobility Analysis and Behavioral Change on Campus
By analyzing anonymized campus Wi-Fi connection logs, we study patterns of student mobility and develop predictive models for applications such as cafeteria congestion forecasting. Beyond prediction, this research investigates how analytical insights can be presented to encourage positive behavioral change. The project integrates privacy-preserving data processing with mobility analytics to balance individual privacy and practical campus optimization.

AI Privacy and Security

Topic: Efficient Model Aggregation and Attack Resilience in Federated Learning
Federated Learning (FL) has attracted significant attention as a distributed learning paradigm that enables AI training without sharing raw data. This research focuses on developing efficient model aggregation methods that reduce communication costs while maintaining model accuracy. In addition, we investigate algorithmic defenses to enhance robustness against backdoor attacks and model extraction attacks in federated environments.
Topic: Jailbreak Attacks and Defense Mechanisms for Large Language Models
Large Language Models (LLMs) demonstrate advanced conversational capabilities but remain vulnerable to jailbreak attacks that can induce unintended or unsafe outputs. This research aims to systematically analyze attack mechanisms and develop effective defense strategies to improve the safety and reliability of LLM systems.