The release ceremony for 2022 World Leading Internet Scientific and Technological Achievements, one of the highlights of the 2022 World Internet Conference Wuzhen Summit, was held in Wuzhen town, Zhejiang province on Nov 9, 2022. A total of 15 world-class scientific and technological achievements were unveiled at the summit. Ruijian System and Equipment for Digital Forgery Detection developed by ICT’s Media Synthesis and Forensics Lab was selected among the 15 world-class achievements.
With production and dissemination of false information on the Internet bringing risk to public security, the team from the Media Synthesis and Forensic Lab developed the first Chinese automatic false news detection system Ruijian Shiyao and the first forgery detection system optimized with software and hardware collaboration and equipment. These products make essential contributions to online content security.
Deconstructing technical proble ms for a credible base of data in Web 3.0
Information forgery is a common issue on the global Internet. The team organized a comprehensive picture of false information detection in terms of algorithms, systems and products. After nine years of collaborative work, two milestones were achieved and applied in practice:
1. Development of the first Chinese automatic false news detection system Ruijian Shiyao. To combat information corruption, task uncertainty, and environmental adversary in real-world scenarios, the team proposed a series of methods for information credibility measurement from diverse viewpoints, including from the perspectives of content, the user, and propagation. The platform has been running for 9 years, accumulating millions of controversial news clues and 100,000 data sets of precise standard rumors. It has a role in the battle against false information in important events.
2. Development of equipment for image/video forgery detection and attribution. To satisfy the urgent need for large-scale high-accuracy, high-performance deployment of equipment in real scenarios, the team developed the first Chinese-produced chip-based equipment for forgery detection, based on software and hardware cross-layer optimization technology. With an average detection time of 10 ms per image and an acceleration ratio of 10 times, it supports the concurrent demand of handling network traffic of 2 GB/sec, while greatly lowering the hardware cost for large-scale deployment.
Chinese Automatic False News Detection System Ruijian Shiyao
Deploying as the key project, and forming a closed loop of supervision and governance of false and forged information on the Internet
Products the team developed have been applied by government agencies and major online media platforms to protect information security on the Internet. At recent major international events, the platform quickly produced special analysis reports on false information circulating online, providing a decision-making basis for national emergency response.
To promote global development in detecting false and forged Internet information, the team has organized several academic forums and competitions. The team also released the largest Chinese multimodal dataset for fake news detection, which included a 2020 Fake News Detection effort during COVID-19 Pandemic Competition. The competition attracted wide attention from global academic and industry peers, and 672 teams participated.
Building false and forgery detection tools for everyone
The team developed the applet Ruijian AI, which provides free news credibility measurement and image/video forgery detection services. To date, more than 310,000 users visited the applet. Since 2018, our Toutiao channel AI Shiyao has been launched to provide timely rumor refutation to the public. The channel has released more than 800 rumor refutation articles, with a cumulative reading volume of over 130 million. The team established long-term cooperation with the TV show "Scientific Gossip List" on BRTV, and with "News Verification Group" on Phoenix TV, to provide more in-depth, accurate news services.
Media Synthesis and Forensics Lab (formerly Multimedia Computing Group) was established in 2002. It undertakes research on the key directions of the 14th Five-Year Plan of ICT-CAS. This lab, one of China’s earliest research teams in the area of online false information detection, focuses on R&D of digital content security. To meet the security needs of the nation and society, the laboratory built a three-level research picture of algorithms, systems and applications. The team developed the first Chinese automatic false news detection system AI Shiyao and the first forgery detection system optimized with software and hardware collaboration and equipment. The academic achievements have been applied to industrial scenarios in public safety, epidemic prevention/control, finance, media and other fields. The lab has published more than 40 papers in international conferences and journals on topics such as natural language processing, computer vision, multimedia and artificial intelligence. The largest Chinese multi-modal fake news detection dataset was released by the lab, which has also held data competitions encouraging development of forgery detection.