Research and Advances
Computing Applications East Asia and Oceania Region Special Section: Big Trends

Activities of National Institute of Informatics in Japan

dot-based map of Japan on concentric rings, illustration

The National Institute of Informatics (NII) is a national-level research and services institute focusing on informatics in Japan. NII was formally established in 2000 by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT). Located in downtown Tokyo, approximately 80 permanent researchers conduct computer science research in various fields. In addition, NII provides networking and information services for more than 1,000 research and educational institutions in Japan. This article describes NII’s activities in these two complementary missions: research and services that support research (as illustrated in Figure 1). The service mission of NII has been fulfilled through three widely deployed academic services: the SINET family of high-performance and availability networks; the Gakunin Research Data Management (RDM) Platform, which supports open access to, and secure sharing of, data-driven research results; and cybersecurity services to protect SINET, RDM, and other services such as secure computation. The research mission of NII can be represented by four fundamental computer science projects and two applied research projects. The fundamental research projects are: graph algorithms, formal verification and their applications, machine vision using fluorescence, and engineerable AI. The applied research projects are: detection of fake videos, audio clips, and documents; and CT image AI-based analytics for the COVID-19 pandemic. The three groups of projects, services in support of research, fundamental research, and applied research, will be briefly described in this article.

Figure 1. NII organizational structure.

Back to Top

Academic Services in Support of Research

SINET backbone network service. One of the most impactful IT services provided by NII is SINET6,6 which has a fully redundant 400Gbps backbone network from Hokkaido to Kyushu, as well as a 200Gbps connection to Okinawa. SINET6 provides high-reliability and high-performance backbone network service to more than 1,000 research and educational institutions in Japan. SINET5 was the previous-generation network, which had a 100Gbps backbone, with an automated, very fast (less than 50msec) switch over when needed. Since SINET5 became operational, NII has provided reliable network services through many natural disasters that have hit Japan over the years.

The typical unique features of SINET6 are its high speed, high density, and multiplicity of network services. SINET6 has 70 SINET nodes nationwide and connects them by the world’s fastest 400 gigabit Ethernet (400GE) lines. This makes it easy for the user institutions to introduce highspeed access lines to SINET nodes at low cost. It is safe to say that SINET6 provides them with the world’s fastest end-to-end network as compared to commercial networks and other research networks. SINET6 also provides a variety of network services compared with other networks. It provides both IPv4/IPv6 dual-stack services for open network access and layer-2/3 virtual private network (VPN) services for secure collaborations. SINET users can set up L2VPNs by specifying the destinations and routes in an on-demand manner through a SINET on-demand controller. This on-demand capability is useful for various research fields such as telesurgery experiments that need assured bandwidth, non-compressed 8K video transmissions that need to select routes to obtain enough bandwidth, and so on. NII has stably provided this variety of services over the cutting-edge high-speed devices by effectively developing and debugging new functions.

The development of the SINET family of networks has been led by Shigeo Urushidani, deputy director of NII. In addition to the backbone network, NII also provides Mobile SINET service for research and innovative IoT applications, such as precision agriculture (for example, cattle monitoring) and atmospheric phenomena (for example, polar mirage monitoring). In parallel with the technological evolution of the SINET family of backbone networks, Mobile SINET has evolved from LTE to 5G, with support for private 5G service (see Figure 2).

Figure 2. SINET 6: Academic backbone network in Japan.

Data platform service. NII RCOS (Research Center for Open Science and Data Platform),5 under the direction of Kazutsuna Yamaji, is responsible for the development of the NII Research Data Cloud. The initial version of the NII RDC consists of three primary platforms: a research data management platform (GakuNin RDM), a publication platform (JAIRO Cloud), and a discovery platform (CiNii Research). These platforms aim to be utilized along with the research data life cycle. GakuNin RDM is currently used by more than 60 research institutions. It aims to be a vital service that enables researchers to share data between collaborators in their daily activities. In addition to data sharing, GakuNin RDM also supports the Data Management Plan enforced by the funding agency, metadata management, and a data analysis environment with appropriate secure storage capacity. Reliable research output managed by GakuNin RDM is publicized from the cloud service of institutional repositories named JAIRO Cloud, which is employed by more than 750 universities in Japan. The whole contents stored in each repository are aggregated by CiNii Research which has an enormous academic knowledge graph in Japan.

In addition to the symbolic European Open Science Cloud, similar activity can be seen globally, such as Australian Research Data Commons, Korean Research Data Platform, Malaysia Open Science Platform,7 and African Open Science Platform. NII RDC continues its effort to build interoperability with these platforms. Compared with them, the advantage of NII RDC is the national-scale deployment and comprehensive service through the research data life cycle. The National Research and Education Network (NREN) in each country provides a high-speed network and several middleware services, including identity and access management federation. The uniqueness of NII is to add value through the research data infrastructure on top of these existing services.

In 2020, the Japanese Cabinet Office recommended researchers to use NII RDC in large-scale publicly funded research projects, which is developing good use cases in practice. For instance, in the Moonshot Goal 2 project in Japan realization of ultra-early disease prediction and intervention by 2050, life scientists and theoretical scientists share and manage research data using GakuNin RDM. They also utilize data analysis tools belonging to GakuNin RDM and share analysis programs and mathematical models. The program aims to publicize their research data using JAIRO Cloud. NII RDC has become a vital research infrastructure in their research program. In 2022, Japan’s MEXT started a new program for nationwide and institutional wide deployment of GakuNin RDM. Through this program, the laboratory in each university will get an opportunity to be in touch with GakuNin RDM and gain its benefits.

NII RDC is not only to provide a feasible research environment but also to change the conservative research culture to facilitate the data-centric science. As illustrative example, visualizing the adoption of a standard dataset by many research papers will increase the recognition of its impact. By adding a new dimension to the current exclusive emphasis on paper citation counts, scientists who collect and publish high-quality datasets will also become highly regarded (see Figure 3).

Figure 3. Architecture of NII Research Data Cloud.

Cybersecurity service. Under the direction of Hiroki Takakura, the Center for Strategic Cyber Resilience R&D coordinates and supports NII Security Operation Collaboration Services (NII-SOCS), which operates security monitoring and provides technical advice for national universities in Japan. In addition to countering and mitigating overt cyberattacks, the center implements technologies for detecting signs of advanced persistent threats (APT) and deploying countermeasures to block the APT and their follow-up attacks such as ransomware.

With the advent of COVID-19, our lifestyles have been drastically transformed. This transformation also affected communications between SINET and the Internet, which are monitored by NII-SOCS. It is natural for everyone to take full advantage of the convenience of cloud computing, and in addition to legitimate users, attackers also use major cloud service providers. More than 90% of communications are now encrypted. As the results, traditional security measures, including pattern-matching, indicators of compromise, are rapidly losing their effectiveness.

Under such circumstances, we developed various analysis methods, for example, the combination of threat intelligence analysis and parallel processing by GPUs to find malicious activities in communications at several hundred Gbps.2 As a result, we have succeeded in narrowing down the billions of suspicious communications generated daily to a few dozen high-risk ones. NII-SOCS also analyzes a variety of threat information and requests enhanced defenses when vulnerabilities are discovered in the networks of participating universities. This activity contributes to the prevention of damage such as unauthorized intrusion into VPN routers, which has been recently occurring around the world.

NII-SOCS fully utilizes the project of SINET 6, traffic mirroring functions. In this service, the traffic of their access lines is mirrored and forwarded to specified hub locations with security analysis devices. Since the amount of mirrored traffic is enormous, the target for mirroring is dynamically selected and turned on or off.

To stimulate research activities on cybersecurity, the center provides research data, including sanitized traffic benchmark and malware data with analysis reports based on observed attacks.

The center also conducts table-top exercises, role-based discussion groups with scenarios modeled on actual cyberattacks that have occurred recently in universities and other organizations. These discussion-based simulated exercises modeled on actual cyberattack scenarios are designed for all enterprise personnel, including technical and administrative staff, as well as executives. These simulated exercises help universities build resilience, to continue their operations even in the face of cyberattacks. Through these activities, the center contributes to the broadening of cybersecurity expertise through the entire enterprise, providing a potential remedy for the global shortage of dedicated cybersecurity professionals.

Back to Top

Fundamental Research in Computer Science

Graph algorithms and challenge for minimum cut(mincut). Ken-ichi Kawarabayashi shared Fulkerson Prize in 2021 with Mikkel Thorup of the University of Copenhagen, the highest honor in discrete mathematics and algorithms, for their deterministic mincut algorithm for a simple graph. Mincut is one of the central problems in graph algorithms, starting from the seminal algorithm of Ford and Fulkerson (1956). It has been still a challenge though, to extend to multiple graphs.

Formal verification and applications. Ichiro Hasuo’s group made significant contributions to software science, ranging from mathematical foundations to real-world applications. In their recent work3 on safety of automated driving vehicles (ADVs), they introduced the first formal logic that proves safety in the pull-over scenario (Figure 4). The scenario, while essential for level 4–5 ADVs, poses a unique challenge due to its complexity. Their logic overcame the complexity by compositional reasoning, a central paradigm of deductive verification à la Antony Hoare.

Figure 4. The pull-over scenario for autonomous driving.

Innovative computer vision based on fluorescence and photoacoustic signals. Imari Sato’s pioneering work analyzing reflective-fluorescence images has opened a new research field in computer vision.7 Based on the unique properties of fluorescence, she presented a series of solutions to challenging computer vision problems, including light color estimation, robust shape estimation of concave objects, and liquid classification. She has also made important achievements in photoacoustic imaging technology, a state-of-the-art noninvasive measurement technique, and realized a fully automated system for 3D visualization of blood and lymphatic vessels (see Figure 5).

Figure 5. Photoacoustic imaging for vessels.

Engineerable AI. Fuyuki Ishikawa of NII is leading the “Engineerable AI (eAI)”a project for trustworthy AI engineering. One of the notable outcomes is techniques for targeted and regression-controlled update of neural networks by adapting fault localization techniques, originally used for estimating the bug locations. These techniques were tested by benchmarks defined with the automotive industry and demonstrated the capability to adjust and improve prediction performance of state-of-the-art prediction AI to meet fine-grained safety requirements for automated driving (see Figure 6).

Figure 6. Engineerable AI for autonomous driving.

Back to Top

Applied Research for Societal Benefits

Detection of fake images, audio clips, and documents. During the COVID-19 pandemic, fake news items have caused significant harm, a problem recognized as an “infodemic” by the World Health Organization (WHO). Attackers can generate and distribute fake videos, fake audio clips, and fake documents for purposes of fraud, propaganda, and public opinion manipulation, using widely available tools such as the DeepFake1 image generator. To detect fake media, Isao Echizen and his colleagues founded the Global Research Center for Synthetic Media (SynMedia Center). The SynMedia Center conducts research and development to generate and detect synthetic and fake media, ensure media reliability, and support decision making. A concrete example is the AI-based SYNTHETIQ VISION tool (September 2021) that automatically detects fake facial videos generated by tools such as DeepFake. This fake detector program is available as a Web API and illustrates “AI as a service (AIaaS).”

AI-based COVID-19 pneumonia diagnosis tool using CT image analytics.b When the COVID-19 pandemic began in early 2020, PCR (polymerase chain reaction) testing stations were relatively rare in Japan. For more serious cases that required hospitalization, NII and the Japan Radiological Society developed an AI-based COVID-19 pneumonia diagnosis tool, using computerized tomography (CT) images. Although the data volume of each CT image set is very large, we were able to leverage SINET6 to collect more than 700 sets of CT images of COVID-19 pneumonia cases from around Japan. The AI-based diagnostic tool was developed quickly and achieved a detection accuracy of 83%. This achievement was reported widely, including in an article in MIT Technology Review (July 2021). This system was one of the earliest AI-based COVID-19 pneumonia diagnosis detectors in the world. Kensaku Mori of Nagoya University played a leading role in the NII development team, in collaboration with medical imaging researchers from the University of Tokyo, Kyushu University and others, achieving success through Team Science (see Figure 7).

Figure 7. Medical imaging big data platform built by NII.

Back to Top


A growing number of successful NII projects have built on vigorous international collaborations. For example, the 2021 Fulkerson Prize was given in recognition for a paper by Kawarabayashi and Thorup of the University of Copenhagen. Beyond the illustrative projects described in this article, the NII team sincerely wishes to expand global collaborations in many areas of computer science research.


    1. Afchar, V. et al. MesoNet: A compact facial video forgery detection network. In Proceedings of the 2018 IEEE Intern. Workshop on Info. Forensics and Security, 1–7; doi:10.1109/WIFS.2018.8630761.

    2. Ando, R., Kadobayashi, Y. and Takakura, H. Choice of parallelism: Multi- GPU driven pipeline for huge academic backbone network. Intern. J. Parallel, Emergent and Distributed Systems 36, 4 (2021), 6.

    3. Hasuo, I. et al. Goal-aware RSS for complex scenarios via program logic. IEEE Trans. Intelligent Vehicles, 2022; doi:10.1109/TIV.2022.3169762.

    4. Kawarabayashi, K. and Thorup, M. Deterministic edge connectivity in near-linear time. JACM 66, 1 (Dec. 2018). Article 4, 1–50,

    5. Komiyama, Y. and Yamaji, K. Nationwide research data management service of Japan in the open science era. In Proceedings of the 6th IIAI Intern. Congress on Advanced Applied Informatics, 2017, 129–133;

    6. Kurimoto, T. et al. SINET6: Nationwide 400GE-based academic backbone network in Japan. To appear in OFC2023.

    7. Zhang, C. and Sato, I. Image-based separation of reflective and fluorescent components using illumination variant and invariant color. IEEE Trans. Pattern Analysis and Machine Intelligence 35, 12 (2013), 2866–2877.

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More