Miyaji Laboratory

Participants:He Binchang, Jichao Xiong, Jin Ashida, Shinya Ochiai, Atsuko Miyaji

The 75th Joint Meeting of the IEICE Technical Committees on Information and Communication System Security (ICSS) and Internet Architecture (IA) was held in a hybrid format at e-Topia Kagawa in Takamatsu, Kagawa, on June 17-18, 2026.

The meeting featured 26 technical presentations across five ICSS sessions and three IA sessions. The program covered a broad range of topics, including AI-assisted vulnerability detection and remediation, software bills of materials (SBOMs), cyber-incident response, network reliability, privacy-preserving machine learning, lightweight cryptography, IoT security, anomaly detection, Internet measurement, and software supply-chain attacks. Presentations were given by researchers from universities, companies, and public research institutes, making the meeting a valuable opportunity to examine both academic advances and practical security challenges.

Miyaji Laboratory contributed four presentations on coupled adaptive batch sizes for one-pass locally differentially private SGD, TFHE-based homomorphic inference for continuous behavioral authentication, membership-inference vulnerabilities in label- and feature-protected data, and improved differential-linear analysis of the lightweight block cipher SPECK. These presentations reflected the laboratory's broad research activities spanning privacy-preserving machine learning, fully homomorphic encryption, privacy attacks, and cryptanalysis.

Overall, many studies placed strong emphasis on practical deployment and operational constraints rather than evaluating security techniques only in idealized settings. In particular, the use of AI for automated vulnerability management, the integration of privacy-enhancing technologies into machine-learning systems, and the security of software and network infrastructures were recurring themes. The joint format also highlighted the increasingly close relationship between system security and Internet architecture. The presentations and discussions provided useful perspectives for refining threat models, evaluation methodologies, and the practical relevance of future research.