Since the beginning of the year, a number of image generation software have rapidly become popular all over the world, and the fidelity of their paintings is "astonishing". AI has been "hurricane" all the way, making the threshold for multiple forms of image forgery such as generation and tampering lower, and the hidden dangers caused by this are also worrying.
Images are one of the main carriers of information. Using AI to falsify images is a key issue that needs to be addressed in the development of AI credibility. At the World Artificial Intelligence Conference (WAIC 2023), the China Academy of Information and Communications Technology (CAICT for short) held a special forum on the theme of "Trusted AI with multi-modal basic models". At the forum, Hehe Information's AI image content security technology solution received widespread attention. This solution can accurately locate traces of tampering in screenshots, intelligently distinguish generative images, and prevent criminals from using technical means to illegally crawl information on images.
Caption: Tang Qi, general manager of Hehe Information's Intelligent Innovation Division, introduced the application of Hehe Information's AI image security technology on the spot
Using AI to protect the security of AI image content, the new technology of Hehe Information is very "practical"
At the artificial intelligence conference last year, Hehe Information demonstrated the "PS tamper detection" technology for the first time. On the one hand, AI is really "eyes of fire".
Caption: Hehe Information's AI image tampering detection technology application effect display
This year, the "black technology" of AI image tampering detection has been continuously optimized and upgraded, and its application has also been expanded to "screenshot tampering detection". Compared with last year's natural scene images, screenshots such as qualification certificates, document contracts, and bank insurance policies that have a greater impact on users' lives will suffer incalculable property losses if they are used by criminals.
It is understood that the upgraded AI image tampering detection technology can detect various screenshots including transfer records, transaction records, chat records, etc., whether it is "cutting out" key elements from the original image and moving "paste" to another place. Whether it is a method of image tampering such as copying and moving, or methods such as "erasing" and "reprinting", image tampering detection technology can all recognize fakes with "intelligent eyes".
Caption: "Reprint" tampered picture detection example: Input a picture into Hehe Information AI image tampering detection model, and the model will be able to judge whether the image has been tampered with, and locate the tampered area of the image
The main difficulty of this technology is that, compared with certificate tampering detection, the background of the screenshot has no texture and background color, and the entire screenshot has no lighting difference. Mining the fine-grained differential features of original images and doctored images.
To this end, Hehe Information proposes an image authenticity identification model based on the encoder-decoder structure of HRNet, which combines the information of the image itself, including but not limited to noise, spectrum, etc., so as to capture fine-grained visual differences. Achieve high-precision identification effect.
In response to the AIGC topic that is "hot searched" on the frequency, Hehe Information has developed an AI-generated picture identification technology to judge whether the picture is generated by AI. One of the difficulties in image recognition generated by AIGC is that there are many image scenes generated by AI, which are difficult for machines to distinguish. Combined information is based on spatial domain and frequency domain relationship modeling, and can use multi-dimensional features to distinguish subtle differences between real pictures and generated pictures.
In addition, Hehe Information has carried out innovative research on OCR anti-attack technology. When people send pictures containing personal information on the Internet, the information contained in these materials is likely to be extracted and leaked by criminals using OCR technology. OCR anti-attack technology can disturb the text of the scene or the text in the document, "lock" the important files of individuals or enterprises, prevent third parties from reading and saving all the text content in the image through the OCR system, and reduce the risk of data leakage. To achieve the purpose of protecting private information.
Caption: Hehe Information’s OCR anti-attack technology application effect display, which can cover key information and prevent machines from automatically crawling without affecting the naked eye
Promote the establishment of industry development standards for image content security , and help the systemic implementation of trusted AI
The "White Paper on Artificial Intelligence (2022)" issued by the China Academy of Information and Communications Technology mentioned that artificial intelligence is entering a new stage of "innovation-driven, application-deepening, and standardized development". The realization of trusted AI requires the coordination of various forces such as the government, enterprises, and universities. The establishment of industry standards may provide strong support for China's "trusted AI" system construction in the fields of machine vision and image processing.
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 Measures (Draft for Comment)" and other documents, and systematically establish the development order of the image content security industry, in June this year, China Academy of Information and Communications Technology took the lead in launching the formulation of the "Document Image Tampering Detection Standard", and Hehe Information, China Image and Graphics Society, University of Science and Technology of China and other scientific and technological innovation companies and well-known academic institutions jointly participated in the preparation.
It is reported that the standard will be based on the current status 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. Tang Qi, general manager of the Intelligent Innovation Department of Hehe Information, said in the forum sharing that the launch of the "Document Image Tampering Detection Standard" formulation project is an important milestone on the road to the construction of an AI image content security system. We look forward to the participation of more research institutions and business entities to jointly promote the overall improvement of the standardization of AI services and the establishment of an industry development order, so as to help science and technology move forward and develop for good.
The healthy development of trusted AI requires not only standard guidance, but also "hard technology" support. Hehe Information has been deeply involved in the fields of intelligent text recognition and intelligent image processing, and its technical achievements have been recognized by authoritative organizations and the market. In the last month, the company's intelligent document processing products passed the evaluation of "Trusted AI-Intelligent Document Processing System" of China Academy of Information and Communications Technology, and obtained the "Level 5" assessment. "Level 5" is the highest rating for this module.
At present, Hehe Information's AI technology has been implemented in nearly 30 industries such as banking, insurance, securities, manufacturing, and government affairs, serving individual and corporate users in more than 200 countries and regions around the world. In the future, Hehe Information will work with academic and industrial partners to help the safe and credible development of the AI image content industry, and create a bright business card for the development of artificial intelligence in Shanghai.