Research on data management, query processing and system automation. Recently, we have carried out research including the following.
- Spatio-temporal data management and query processing
- Distributed system
- Automated system tuning
- Machine learning for database
Organizations collect vast amounts of information on individuals. At the same time, they have access to increasing levels of computational power. Although this conjunction of information and power provides great benefits to society, it also threatens individuals’ privacy. Thus, data privacy and protecting individuals’ anonymity have become mainstream for research. Recently, we have carried out research including the following.
- Differential privacy
- Privacy-preserving maching learning
- Federated learning
- Machine unlearning
Health data may contain information that can identify specific individuals, which can lead to privacy violations during collection. Therefore, there is a need for data processing methods that protect privacy when collecting health data while maintaining usability of the data so that tasks such as arrhythmia classification are still possible. Not only can it be difficult to collect data due to privacy issues, but data bias can also occur due to the small number of data from unhealthy people compared to healthy people. Therefore, data generation methods that can solve the problems of data insufficiency and bias models are needed. Recently, we have carried out research including the following.
- Privacy-preserving sharing of health data
- Generative adversarial networks