Taiwan's First PETs Frontier Cybersecurity Technology Forum
With the use of cloud services, 5G, IoT, and other cybertechnology becoming increasingly widespread, a data streaming society is within reach. Since big data usage is also a major trend, scientists must strive to take into account public interest while also enhancing the protection of personal data; this is an important scientific research topic that the Ministry of Science and Technology (MOST) is committed to promoting.
In order to provide high accessibility and reliable privacy protection for cloud data, NARLabs' National Center for High‑Performance Computing (NCHC) under MOST has not only accelerated the development and introduction of a new generation of personal data protection technology in Taiwan by utilizing national high-speed computing facilities, but also held Taiwan's first "Privacy Enhancing Technologies (PETs) Frontier Cybersecurity Technology Forum" on October 15th. The forum gathered experts in the fields of encryption and privacy protection technology for in-depth discussions with the hope of accelerating the achievement of Taiwan's vision of "Digital Nation, Smart Island." Participating institutions from around the world included Academia Sinica, National Taiwan University, Yang Ming Chiao Tung University, IBM, Duality Technologies, and Nvidia.
The NCHC is implementing a frontier digital infrastructure project to build a fleet of Taiwania supercomputers which will provide world-class supercomputing and high-speed AI computing services to domestic industry, academia, and research institutes. In order to strengthen protection of the nation's digital sovereignty in the future, the development of domestic PET-related technologies such as homomorphic encryption and federated learning is also being promoted.
Taking homomorphic encryption as an example, in the past, obtaining computation results when uploading encrypted data to the cloud and directly operating on the encrypted text took millions of times longer as it required a great amount of computing resources. However, with increased computing power and algorithmic breakthroughs in recent years, this year's testing of the Taiwania supercomputer by the NCHC has allowed acceleration to reach a level nearing practical application, verifying that this new method is beginning to show possibility for being put into practice.
The NCHC has also newly made public its self-developed homomorphic encryption system based on quantum computing concepts. The number of incorrect choices added to the encryption is exponentially greater, allowing it to resist attacks from quantum computers. Moreover, through the transformation of a large number of quantum gates, errors will not be accumulated in the computation process, no matter how long it is. Unlike traditional homomorphic encryption technology, which inherently accumulates errors quickly, the new system is faster and more accurate. In addition, the huge number of quantum gates in the computation process allows for proper permutation; therefore, the original mode of computation will remain unknown to observers, and data can be masked when computation is performed in the cloud. This approach to quantum concept computing completely solves the predicaments of existing technology and realizes the possibility of homomorphic encryption, which not only helps open up new research directions, but also may create new models for industry.
Federated learning also contributes to privacy protection by providing a decentralized way for data to be interoperable through AI training models without leaving its location, and the training results for all data can be shared. The NCHC has successfully developed a program integrating Nvidia Clara to design a browser user interface that lowers the threshold for using federated learning.
NCHC Director-General Shepherd Shi indicated that by using the NCHC's supercomputers as an analog verification environment, experimentation for more frontier privacy enhancing technologies, including new encryption technology that can be adapted to the post-quantum era, can be offered. These can be beneficial to both development and the public interest of personal data privacy and protection, and are indeed significant benefits of national digital infrastructure.