Document images are an important carrier of information, but they are often tampered with by criminals using software and algorithms. These false materials are often used to spread rumors, economic fraud, and fabricate false news, which have caused bad effects on individuals and society. After the global explosion of AIGC, people are worried about the rising risk of "generative counterfeiting", and the safety and credibility of image content has also become the focus of public attention.
In order to implement the normative requirements for AI services in the "Network Security Law of the People's Republic of China", "Generative Artificial Intelligence Service Management" and other documents, and systematically establish the development order of the image content security industry, the China Academy of Information and Communications Technology (hereinafter referred to as "China ICT Institute") initiated the formulation of the "Document Image Tampering Detection Standard". The standard is led by China Academy of Information and Communications Technology, and jointly compiled by Shanghai Hehe Information Technology Co., Ltd., China Image and Graphics Society, University of Science and Technology of China and other scientific and technological innovation enterprises and well-known academic institutions.
The "Document Image Tampering Detection Standard" will provide a reliable guarantee for the security of document image content and help the construction of an AI security system in the new era. As the lead party, China Academy of Information and Communications Technology stated that the "Document Image Tampering Detection Standard" will be based on the status quo of the industry, focusing on "fine-grained" visual difference forgery image identification, generative image identification, document image integrity protection and other industry focus issues to gather industry consensus , in order to provide effective guidance for the industry.
The risk of image tampering is imminent, and the introduction of standards is imminent
With the development of technology and the increase in the penetration rate of electronic equipment, the threshold for people to modify images has become lower and lower. In the past, image content processing required the use of professional software. Now through certain platforms or software, it can be produced by moving fingers. A false picture, the originality and authenticity of digital images are facing serious challenges.
Previously, the technical research objects of image tampering detection mainly focused on natural scene images. However, it is usually tampered qualification certificates, documents, chat screenshots, etc. that really bring risks to people's lives. It is often difficult for traditional detection systems to judge the tampering of these document images and locate the modified content. The main reasons are as follows:
On the one hand, the variety of tampering methods in real scenes and the subtle visual traces of tampering pose challenges to the tampering detection scheme; the complexity of document layout and the diversification of text content have also become "blockers" in the process of image content detection and recognition, increasing the This increases the difficulty of implementing tamper detection.
In addition, when the traditional image tampering detection method is faced with global cropping, toning processing and combined forgery techniques, there is still room for improvement in its detection coverage, accuracy, and security, which requires multiple parties such as the government, research institutions, and enterprises. Join together to help the healthy growth of the AI visual image industry by establishing technical specifications and standards.
Image content security has become the key construction direction of "trusted AI" in the future
AI is a "double-edged sword". The "Report on the Ecological Development of the Trusted Artificial Intelligence Industry (2022)" issued by the China Academy of Information and Communications Technology mentioned that trustworthiness has become an essential element for the development of the artificial intelligence industry, driving the normative development of artificial intelligence.
"Trustworthy" reflects the trustworthiness of AI systems, products, and services in a series of inherent attributes such as security, reliability, explainability, and accountability. Related concepts are changing from vague to clear with the pace of technological development. From beginning to maturity. Since the beginning of the year, AIGC products represented by language generation and visual generation models have triggered social discussions on the "credibility" of graphic content. The formulation of the "Document Image Tampering Detection Standard" may provide strong support for China's "trusted AI" system construction in the fields of machine vision and image processing.
The use of scientific and technological means to achieve accurate detection of text tampering marks will provide a reliable guarantee for the security of document image content and help the construction of an AI security system in the new era. In the field of image content security, some companies have launched cutting-edge explorations, using scientific and technological means to implement authenticity judgments and tampering trace detection on tampered text images. Artificial intelligence technology company Hehe Information is a representative of them.
It is reported that Hehe Information has developed image tampering detection technology and related systems based on deep learning, which can detect various forms of tampering, intelligently capture the subtle traces left by the image during the tampering process, and display image area tampering in the form of a heat map Location, related technologies have been applied in banking, insurance and other industries.
According to the public information, Hehe Information focuses on the research of the underlying technology and application products of intelligent document processing. The related technology has won the championship in more than ten competitions such as ICPR and ICFHR, and won the championship in CVPR, AAAI, ACL, ACM MM, etc. Published at the International Summit. At present, the company has obtained more than 100 invention patents with independent intellectual property rights, and provides intelligent text recognition products and AI services for hundreds of millions of users in more than 200 countries and regions around the world.