Fujitsu to combat fake news in collaboration with leading Japanese organizations

Fujitsu announced that it has begun a project to develop a disinformation countermeasure platform alongside a consortium of leading academic and private sector organizations. Fujitsu was selected as a primary operator for this initiative in July 2024 through a public call for proposals by Japan’s New Energy and Industrial Technology Development Organization (“NEDO”). The effort is part of the Key and Advanced Technology R&D through Cross Community Collaboration Program (“K Program”), which was established with the collabor”ation of Japan’s Cabinet Office, the Japanese Ministry of Economy, Trade and Industry (METI), and other related Japanese ministries, to strengthen and drive Japan’s economic security. The consortium, assembled by Fujitsu, includes the National Institute of Informatics (NII), NEC Corporation, Keio Research Institute at SFC, Institute of Science Tokyo (formerly Tokyo Tech), The University of Tokyo, University of Aizu, Nagoya Institute of Technology, and Osaka University.

The project aims to develop the world’s first comprehensive disinformation countermeasure platform that can process false information from initial detection to evidence gathering, analysis, and evaluation, with development slated for completion by the end of fiscal year 2025.

Vivek Mahajan, Corporate Vice President, CTO, CPO, Fujitsu Limited, comments: “We are excited to be working on this initiative with a top-tier consortium of Japanese academic and private sector organizations that have a proven track record of combating disinformation. In addition to our consortium partners, Fujitsu will collaborate with relevant government agencies and other organizations to develop a robust countermeasure solution and contribute to solving this serious societal challenge.”

As prime operator of this project, Fujitsu will lead the R&D efforts, technology integration, and construction of the overall platform. Leveraging the latest trends in disinformation tactics and technologies, Fujitsu will also create use cases for public and private sector organizations, facilitating the practical application of the research findings.

Technology to be developed and consortium roles

1. Information analysis by media type and disinformation detection (Responsible parties: NII, NEC)

NII

Leveraging its track record in the field of fake media detection, NII will develop a technology to detect deepfakes of images, video and audio from social media posts and other content. The technology will identify how the media content has been created and any areas that have been manipulated and assign a confidence score which will be provided as supporting evidence.

NEC

NEC will develop media understanding technology that extracts content including images, video and audio as text and uses it to analyze matches with social media posts and to collect supporting information.

2. Evidence/endorsement management (Responsible parties: Keio Research Institute at SFC, Fujitsu, Osaka University)

Keio Research Institute at SFC and Fujitsu

Keio Research Institute at SFC which has a track record in the field of trusted internet architecture and Fujitsu develop technology to integrate evidence collected from the internet, including the analysis results from technology 1 above. This evidence will be structured and stored as an evidence/endorsement graph (3) to verify authenticity and assess impact.

Osaka University

Osaka University will leverage its expertise in IoT data analysis to develop technology to collect sensor data as a source of evidence. In cases where information for the target area is incomplete, the system will infer evidence based on available data from neighboring areas. This proactive, AI-driven evidence gathering, mimicking human information collection and inference, is a highly advanced initiative.

3. Comprehensive authenticity determination (Responsible parties: Fujitsu, Nagoya Institute of Technology)

Fujitsu

– Fujitsu will develop technology to analyze the consistency of evidence linked to target information using evidence/endorsement graphs from technology 2. This technology enables comprehensive information authenticity analysis, presenting the results and supporting evidence in a user-friendly manner.

– Fujitsu will also develop a specialized Japanese LLM for disinformation countermeasures, utilizing its supercomputer (Fugaku) and LLM (Takane) expertise. This specialized LLM will enhance comprehension of news and social media data, improve logical reasoning capabilities, and enable high-speed, high-accuracy inference for authenticity verification while mitigating hallucinations.

Nagoya Institute of Technology

Fujitsu will collaborate with Nagoya Institute of Technology to develop a user interface and information provision technology based on cognitive science, considering factors related to human psychology (e.g., continued influence effect). This will help users to accurately judge the truthfulness of information and encourage appropriate actions, including making efforts to avoid the unintentional spread of false information.

4. Evaluation of disinformation impact (Responsible parties: Institute of Science Tokyo, The University of Tokyo, Aizu University)

Institute of Science Tokyo, The University of Tokyo, and University of Aizu, with their expertise in computational social science related to social media, will develop technology to assess the impact of disinformation. By extending LLMs to build an AI model for disinformation assessment, the three parties will analyze characteristics of disinformation, such as similarity to past disinformation and proliferation speed, focusing on the source, content, and social context of messages from social media data. This will allow for the evaluation of metrics such as proliferation scale and social impact. This development of technology to visualize and quantitatively assess the social impact of disinformation represents a highly advanced endeavor.